In [1]:
import warnings
warnings.filterwarnings("ignore")
import pandas as pd
import scipy.stats as stats
import numpy as np
from pandas.api.types import is_numeric_dtype
import statsmodels.api as sm 
import seaborn as sns
from seaborn_qqplot import QQPlot
from matplotlib import pyplot as plt
from scipy.stats import gamma
from sklearn.preprocessing import StandardScaler
import pickle 
from sklearn.metrics import mean_squared_error
#-------Importing tensorflow libraries-------#
from tensorflow.python.keras.models import Sequential
from tensorflow.python.keras.layers import Dense,Dropout
from tensorflow.python.keras.wrappers.scikit_learn import KerasRegressor
#------split and pipeline libraries------#
from sklearn.model_selection import train_test_split
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import KFold
from sklearn.pipeline import Pipeline
#--------------------------------------------------#
from statsmodels.stats.outliers_influence import variance_inflation_factor 
sns.set(color_codes=True)
%matplotlib inline
In [2]:
df  = pd.read_csv("Part- 1,2&3 - Signal.csv")
In [3]:
df.head()
Out[3]:
Parameter 1 Parameter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 Parameter 7 Parameter 8 Parameter 9 Parameter 10 Parameter 11 Signal_Strength
0 7.4 0.70 0.00 1.9 0.076 11.0 34.0 0.9978 3.51 0.56 9.4 5
1 7.8 0.88 0.00 2.6 0.098 25.0 67.0 0.9968 3.20 0.68 9.8 5
2 7.8 0.76 0.04 2.3 0.092 15.0 54.0 0.9970 3.26 0.65 9.8 5
3 11.2 0.28 0.56 1.9 0.075 17.0 60.0 0.9980 3.16 0.58 9.8 6
4 7.4 0.70 0.00 1.9 0.076 11.0 34.0 0.9978 3.51 0.56 9.4 5
In [4]:
df.isnull().any()
Out[4]:
Parameter 1        False
Parameter 2        False
Parameter 3        False
Parameter 4        False
Parameter 5        False
Parameter 6        False
Parameter 7        False
Parameter 8        False
Parameter 9        False
Parameter 10       False
Parameter 11       False
Signal_Strength    False
dtype: bool
In [5]:
df.info(verbose=True)

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1599 entries, 0 to 1598
Data columns (total 12 columns):
Parameter 1        1599 non-null float64
Parameter 2        1599 non-null float64
Parameter 3        1599 non-null float64
Parameter 4        1599 non-null float64
Parameter 5        1599 non-null float64
Parameter 6        1599 non-null float64
Parameter 7        1599 non-null float64
Parameter 8        1599 non-null float64
Parameter 9        1599 non-null float64
Parameter 10       1599 non-null float64
Parameter 11       1599 non-null float64
Signal_Strength    1599 non-null int64
dtypes: float64(11), int64(1)
memory usage: 150.0 KB

Converting signal strength in categorical variable

In [6]:
df['Signal_Strength'] = df['Signal_Strength'].astype('object')
In [7]:
df.info(verbose=True)

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1599 entries, 0 to 1598
Data columns (total 12 columns):
Parameter 1        1599 non-null float64
Parameter 2        1599 non-null float64
Parameter 3        1599 non-null float64
Parameter 4        1599 non-null float64
Parameter 5        1599 non-null float64
Parameter 6        1599 non-null float64
Parameter 7        1599 non-null float64
Parameter 8        1599 non-null float64
Parameter 9        1599 non-null float64
Parameter 10       1599 non-null float64
Parameter 11       1599 non-null float64
Signal_Strength    1599 non-null object
dtypes: float64(11), object(1)
memory usage: 150.0+ KB

Univariate Analysis

In [8]:
df.describe()
Out[8]:
Parameter 1 Parameter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 Parameter 7 Parameter 8 Parameter 9 Parameter 10 Parameter 11
count 1599.000000 1599.000000 1599.000000 1599.000000 1599.000000 1599.000000 1599.000000 1599.000000 1599.000000 1599.000000 1599.000000
mean 8.319637 0.527821 0.270976 2.538806 0.087467 15.874922 46.467792 0.996747 3.311113 0.658149 10.422983
std 1.741096 0.179060 0.194801 1.409928 0.047065 10.460157 32.895324 0.001887 0.154386 0.169507 1.065668
min 4.600000 0.120000 0.000000 0.900000 0.012000 1.000000 6.000000 0.990070 2.740000 0.330000 8.400000
25% 7.100000 0.390000 0.090000 1.900000 0.070000 7.000000 22.000000 0.995600 3.210000 0.550000 9.500000
50% 7.900000 0.520000 0.260000 2.200000 0.079000 14.000000 38.000000 0.996750 3.310000 0.620000 10.200000
75% 9.200000 0.640000 0.420000 2.600000 0.090000 21.000000 62.000000 0.997835 3.400000 0.730000 11.100000
max 15.900000 1.580000 1.000000 15.500000 0.611000 72.000000 289.000000 1.003690 4.010000 2.000000 14.900000

Distribution plots for each continous parameters

In [9]:
columns = list(df.drop(columns=['Signal_Strength']).columns)
for column in columns:
    plt.figure(figsize=(10, 5))
    sns.distplot(df[column], color = "blue").set_title("Distribution of "+column)

Box plot of each parameter with respect to the signal strength

In [10]:
for column in columns:
    plt.figure(figsize=(10, 5))
    sns.boxplot(x='Signal_Strength',y=column, data = df, hue = 'Signal_Strength')
In [11]:
plt.figure(figsize=(10, 5))
sns.countplot(y='Signal_Strength',hue = 'Signal_Strength', data = df)
Out[11]:
<matplotlib.axes._subplots.AxesSubplot at 0x33b31408>

Calculating median , mode ,skewness and kurtosis

In [12]:
for column in columns:
    print(100*"*")
    print("Mean of "+str(column)+"="+str(df[column].mean()))
    print("Median of "+str(column)+"="+str(df[column].median()))
    print("Mode of "+str(column)+"="+str(df[column].mode()[0]))
    print("Skewness in "+str(column)+"="+str(df[column].skew()))
    print("Excess Kurtosis in "+str(column)+"="+str(df[column].kurtosis()))
    print(100*"*")

****************************************************************************************************
Mean of Parameter 1=8.319637273295838
Median of Parameter 1=7.9
Mode of Parameter 1=7.2
Skewness in Parameter 1=0.9827514413284587
Excess Kurtosis in Parameter 1=1.1321433977276252
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 2=0.5278205128205131
Median of Parameter 2=0.52
Mode of Parameter 2=0.6
Skewness in Parameter 2=0.6715925723840199
Excess Kurtosis in Parameter 2=1.2255422501791422
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 3=0.2709756097560964
Median of Parameter 3=0.26
Mode of Parameter 3=0.0
Skewness in Parameter 3=0.3183372952546368
Excess Kurtosis in Parameter 3=-0.7889975153633966
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 4=2.5388055034396517
Median of Parameter 4=2.2
Mode of Parameter 4=2.0
Skewness in Parameter 4=4.54065542590319
Excess Kurtosis in Parameter 4=28.617595424475443
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 5=0.08746654158849257
Median of Parameter 5=0.079
Mode of Parameter 5=0.08
Skewness in Parameter 5=5.680346571971722
Excess Kurtosis in Parameter 5=41.71578724757661
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 6=15.874921826141339
Median of Parameter 6=14.0
Mode of Parameter 6=6.0
Skewness in Parameter 6=1.250567293314441
Excess Kurtosis in Parameter 6=2.023562045840575
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 7=46.46779237023139
Median of Parameter 7=38.0
Mode of Parameter 7=28.0
Skewness in Parameter 7=1.515531257594554
Excess Kurtosis in Parameter 7=3.8098244878645744
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 8=0.9967466791744833
Median of Parameter 8=0.99675
Mode of Parameter 8=0.9972
Skewness in Parameter 8=0.07128766294945525
Excess Kurtosis in Parameter 8=0.9340790654648083
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 9=3.311113195747343
Median of Parameter 9=3.31
Mode of Parameter 9=3.3
Skewness in Parameter 9=0.19368349811284427
Excess Kurtosis in Parameter 9=0.806942508246574
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 10=0.6581488430268921
Median of Parameter 10=0.62
Mode of Parameter 10=0.6
Skewness in Parameter 10=2.4286723536602945
Excess Kurtosis in Parameter 10=11.720250727147674
****************************************************************************************************
****************************************************************************************************
Mean of Parameter 11=10.422983114446502
Median of Parameter 11=10.2
Mode of Parameter 11=9.5
Skewness in Parameter 11=0.8608288069184189
Excess Kurtosis in Parameter 11=0.20002931143836733
****************************************************************************************************

Observations:

  • Our dataset has outliers as per the boxplot
  • From the above statistics it can be inferred that Parameters 8 and 9 are normally distributed which can be visually confirmed after we go through the distribution plots of the aforementioned parameters.

    • Multivariate Analysis

    In [13]:
    sns.pairplot(df,hue='Signal_Strength')
    Out[13]:
    <seaborn.axisgrid.PairGrid at 0x34cad388>

    Observation:We can see some correlation between parameter 1 and parameter 8 but we need to check the correlation statistically

    Let us impute outliers detected through z-score method with median

    In [332]:
    df1 = df.copy()
    In [333]:
    for column in df1.drop(columns=['Signal_Strength']).columns:
    df1[column] = df1[column].apply(lambda x: x if
                        ((x - df1[column].mean())/df1[column].std()) <= 3 
                        else
                        df1[column].median()
                        #np.nan
                       )

    Correlation Test

    Kbest method with chisquare

    In [334]:
    from sklearn.feature_selection import SelectKBest
from sklearn.feature_selection import chi2
    In [335]:
    df1.shape
    Out[335]:
    (1599, 12)
    In [336]:
    X = df1.drop(columns=['Signal_Strength'])
y = df1['Signal_Strength'].astype('int')
    In [337]:
    orderd_rank_features = SelectKBest(score_func=chi2, k=11)
    In [338]:
    orderd_rank_features.fit(X,y)
    Out[338]:
    SelectKBest(k=11, score_func=<function chi2 at 0x00000000504C93A8>)
    In [339]:
    scores = pd.DataFrame(orderd_rank_features.scores_, columns=['Scores'])
features = pd.DataFrame(X.columns, columns=['Features'])
final = pd.concat([scores,features], axis=1)
    In [340]:
    final.nlargest(11,'Scores')
    Out[340]:
    Scores Features
    6 2761.106769 Parameter 7
    5 120.932364 Parameter 6
    10 43.476995 Parameter 11
    2 13.614614 Parameter 3
    1 12.224059 Parameter 2
    0 9.455325 Parameter 1
    3 7.156913 Parameter 4
    9 5.830052 Parameter 10
    4 0.281285 Parameter 5
    8 0.126300 Parameter 9
    7 0.000214 Parameter 8

    From above method we can find that Parameters 5 , 9 and 8 are least siginficant predictors. We will try to use the below dataset to create a model and check the performance

    In [341]:
    Kbest_features = df1.drop(columns=['Parameter 5','Parameter 9','Parameter 8']).copy()

    Feature importance using ensemble method

    In [342]:
    from sklearn.ensemble import ExtraTreesClassifier
model = ExtraTreesClassifier()
model.fit(X,y)
    Out[342]:
    ExtraTreesClassifier(bootstrap=False, ccp_alpha=0.0, class_weight=None,
                     criterion='gini', max_depth=None, max_features='auto',
                     max_leaf_nodes=None, max_samples=None,
                     min_impurity_decrease=0.0, min_impurity_split=None,
                     min_samples_leaf=1, min_samples_split=2,
                     min_weight_fraction_leaf=0.0, n_estimators=100,
                     n_jobs=None, oob_score=False, random_state=None, verbose=0,
                     warm_start=False)
    In [343]:
    ranked_features = pd.Series(model.feature_importances_,index=X.columns)
ranked_features.nlargest(10).plot(kind='barh')
    Out[343]:
    <matplotlib.axes._subplots.AxesSubplot at 0x72ec7288>

    from above plot we can see that except parameter 11 almost all features are equally important. So we will retain all columns in this case

    Variance Inflation factor

    In [344]:
    X = df.drop(columns=['Signal_Strength'])
vif_data = pd.DataFrame()
vif_data['feature'] = X.columns
vif_data['VIF'] = [variance_inflation_factor(X.values,i) for i in range(len(X.columns))]
    In [345]:
    vif_data
    Out[345]:
    feature VIF
    0 Parameter 1 74.452265
    1 Parameter 2 17.060026
    2 Parameter 3 9.183495
    3 Parameter 4 4.662992
    4 Parameter 5 6.554877
    5 Parameter 6 6.442682
    6 Parameter 7 6.519699
    7 Parameter 8 1479.287209
    8 Parameter 9 1070.967685
    9 Parameter 10 21.590621
    10 Parameter 11 124.394866

    Pairplot after removing outliers

    In [346]:
    sns.pairplot(df1)
    Out[346]:
    <seaborn.axisgrid.PairGrid at 0x72f09548>
    In [347]:
    X = df1.drop(columns=['Signal_Strength'])
vif_data = pd.DataFrame()
vif_data['feature'] = X.columns
vif_data['VIF'] = [variance_inflation_factor(X.values,i) for i in range(len(X.columns))]
    In [348]:
    vif_data
    Out[348]:
    feature VIF
    0 Parameter 1 67.525613
    1 Parameter 2 17.700405
    2 Parameter 3 8.259058
    3 Parameter 4 9.553091
    4 Parameter 5 16.211930
    5 Parameter 6 6.542079
    6 Parameter 7 6.370576
    7 Parameter 8 1331.472877
    8 Parameter 9 960.386214
    9 Parameter 10 29.009573
    10 Parameter 11 133.234695

    Observations:

  • After treating outliers we did not see much improvement in the VIF scores of the parameters.They continue to be very high(>5)
  • In order to solve multicolinearity problem we will use PCA.

    • Standard Scaling

    In [349]:
    scaled = df1.copy()
for column in scaled.drop(columns=['Signal_Strength']).columns:
    scaled[column] = scaled[column].apply(lambda x:
                        (x - scaled[column].mean())/scaled[column].std()
                       )

    Finding the optimal number of principal components that could account for maximum variance in data

    In [350]:
    from sklearn.decomposition import PCA
features = list(scaled.drop(columns=['Signal_Strength']).columns)
pca = PCA(n_components=10)
pca.fit(np.array(scaled[features]))
variance = pca.explained_variance_ratio_
var = np.cumsum(np.round(variance,decimals = 3)*100)
var
    Out[350]:
    array([27.4, 45.1, 59.4, 69.7, 77.9, 84.7, 89.9, 94. , 97. , 99.1])
    In [351]:
    plt.ylabel("% Variance explained")
plt.xlabel("Number of features")
plt.title("PCA analysis")
plt.ylim(20,110)
plt.xlim(0,11)
plt.plot(var)
    Out[351]:
    [<matplotlib.lines.Line2D at 0xfebe1708>]

    Observations: We could pick n_components = 6 as they account for 90% of variance in the dataset

    In [352]:
    pca = PCA(n_components=6)
principalComponents = pca.fit_transform(np.array(scaled[features]))
pcDf = pd.DataFrame(data = principalComponents
             , columns = ['PC1', 'PC2','PC3','PC4','PC5','PC6'])
    In [353]:
    finalDf = pd.concat([pcDf,scaled[['Signal_Strength']]], axis=1)
    In [354]:
    finalDf.info()

    <class 'pandas.core.frame.DataFrame'>
RangeIndex: 1599 entries, 0 to 1598
Data columns (total 7 columns):
PC1                1599 non-null float64
PC2                1599 non-null float64
PC3                1599 non-null float64
PC4                1599 non-null float64
PC5                1599 non-null float64
PC6                1599 non-null float64
Signal_Strength    1599 non-null object
dtypes: float64(6), object(1)
memory usage: 87.6+ KB
    In [355]:
    finalDf['Signal_Strength'].unique()
    Out[355]:
    array([5, 6, 7, 4, 8, 3], dtype=object)

    Split the data

    In [83]:
    from keras.utils.np_utils import to_categorical
categorical_labels = to_categorical(finalDf['Signal_Strength'], num_classes=None)
    In [84]:
    len(categorical_labels[0])
    Out[84]:
    9
    In [85]:
    finalDf
    Out[85]:
    PC1 PC2 PC3 PC4 PC5 PC6 Signal_Strength
    0 1.663006 -1.043868 -1.414160 -0.177092 0.475915 -1.056528 5
    1 0.846262 -2.201520 -0.042708 0.445265 0.145303 0.413887 5
    2 0.742628 -1.364816 -0.714728 0.068164 0.255632 0.095856 5
    3 -2.433041 0.010059 0.731094 -1.526152 -0.687085 -0.277802 6
    4 1.663006 -1.043868 -1.414160 -0.177092 0.475915 -1.056528 5
    ... ... ... ... ... ... ... ...
    1594 2.282332 -0.572072 0.847368 0.008764 0.358428 0.682458 5
    1595 2.437299 0.228083 2.257227 0.697218 0.739969 -0.447815 6
    1596 1.527470 0.345618 1.232464 0.556543 0.799146 -0.068777 6
    1597 2.455420 -0.451180 1.099449 0.390267 1.083614 -0.438698 5
    1598 0.330841 1.277605 1.192040 1.086017 -0.399542 -0.058249 6

    1599 rows × 7 columns

    In [133]:
    xpca = finalDf.drop(columns=['Signal_Strength']).values
ypca = categorical_labels
X_train, X_test, y_train, y_test = train_test_split(xpca, ypca,test_size=0.33, random_state=42)

    Training the neural network on principal components

    In [134]:
    model = Sequential()
model.add(Dense(32, input_dim=6, kernel_initializer='normal', activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(9, activation='softmax'))
model.summary()

    Model: "sequential_8"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_24 (Dense)             (None, 32)                224       
_________________________________________________________________
dropout_16 (Dropout)         (None, 32)                0         
_________________________________________________________________
dense_25 (Dense)             (None, 64)                2112      
_________________________________________________________________
dropout_17 (Dropout)         (None, 64)                0         
_________________________________________________________________
dense_26 (Dense)             (None, 9)                 585       
=================================================================
Total params: 2,921
Trainable params: 2,921
Non-trainable params: 0
_________________________________________________________________
    In [135]:
    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    In [136]:
    history = model.fit(X_train, y_train, epochs=150, batch_size=10,  verbose=1, validation_split=0.2)

    Train on 856 samples, validate on 215 samples
Epoch 1/150
856/856 [==============================] - 0s 403us/sample - loss: 1.8589 - acc: 0.4229 - val_loss: 1.4051 - val_acc: 0.5209
Epoch 2/150
856/856 [==============================] - 0s 120us/sample - loss: 1.3185 - acc: 0.4825 - val_loss: 1.0763 - val_acc: 0.6419
Epoch 3/150
856/856 [==============================] - 0s 124us/sample - loss: 1.1507 - acc: 0.5350 - val_loss: 1.0197 - val_acc: 0.6372
Epoch 4/150
856/856 [==============================] - 0s 120us/sample - loss: 1.1529 - acc: 0.5397 - val_loss: 0.9766 - val_acc: 0.6465
Epoch 5/150
856/856 [==============================] - 0s 182us/sample - loss: 1.1149 - acc: 0.5748 - val_loss: 0.9545 - val_acc: 0.6372
Epoch 6/150
856/856 [==============================] - 0s 132us/sample - loss: 1.1155 - acc: 0.5374 - val_loss: 0.9559 - val_acc: 0.6465
Epoch 7/150
856/856 [==============================] - 0s 204us/sample - loss: 1.0984 - acc: 0.5386 - val_loss: 0.9421 - val_acc: 0.6512
Epoch 8/150
856/856 [==============================] - 0s 124us/sample - loss: 1.0709 - acc: 0.5584 - val_loss: 0.9347 - val_acc: 0.6465
Epoch 9/150
856/856 [==============================] - 0s 123us/sample - loss: 1.0770 - acc: 0.5502 - val_loss: 0.9283 - val_acc: 0.6605
Epoch 10/150
856/856 [==============================] - 0s 119us/sample - loss: 1.0547 - acc: 0.5607 - val_loss: 0.9252 - val_acc: 0.6698
Epoch 11/150
856/856 [==============================] - 0s 131us/sample - loss: 1.0264 - acc: 0.5759 - val_loss: 0.9141 - val_acc: 0.6605
Epoch 12/150
856/856 [==============================] - 0s 122us/sample - loss: 1.0576 - acc: 0.5572 - val_loss: 0.9067 - val_acc: 0.6698
Epoch 13/150
856/856 [==============================] - 0s 133us/sample - loss: 1.0456 - acc: 0.5619 - val_loss: 0.9022 - val_acc: 0.6512
Epoch 14/150
856/856 [==============================] - 0s 120us/sample - loss: 1.0424 - acc: 0.5713 - val_loss: 0.9043 - val_acc: 0.6651
Epoch 15/150
856/856 [==============================] - 0s 134us/sample - loss: 1.0431 - acc: 0.5783 - val_loss: 0.8997 - val_acc: 0.6698
Epoch 16/150
856/856 [==============================] - 0s 126us/sample - loss: 1.0344 - acc: 0.5409 - val_loss: 0.8981 - val_acc: 0.6605
Epoch 17/150
856/856 [==============================] - 0s 122us/sample - loss: 1.0497 - acc: 0.5526 - val_loss: 0.9017 - val_acc: 0.6698
Epoch 18/150
856/856 [==============================] - 0s 117us/sample - loss: 1.0043 - acc: 0.5724 - val_loss: 0.8937 - val_acc: 0.6744
Epoch 19/150
856/856 [==============================] - 0s 118us/sample - loss: 1.0359 - acc: 0.5561 - val_loss: 0.8926 - val_acc: 0.6558
Epoch 20/150
856/856 [==============================] - 0s 133us/sample - loss: 1.0048 - acc: 0.5876 - val_loss: 0.8882 - val_acc: 0.6791
Epoch 21/150
856/856 [==============================] - 0s 126us/sample - loss: 1.0228 - acc: 0.5643 - val_loss: 0.8832 - val_acc: 0.6651
Epoch 22/150
856/856 [==============================] - 0s 131us/sample - loss: 1.0146 - acc: 0.5841 - val_loss: 0.8959 - val_acc: 0.6558
Epoch 23/150
856/856 [==============================] - 0s 126us/sample - loss: 0.9991 - acc: 0.5935 - val_loss: 0.8864 - val_acc: 0.6651
Epoch 24/150
856/856 [==============================] - 0s 126us/sample - loss: 0.9899 - acc: 0.5841 - val_loss: 0.8826 - val_acc: 0.6605
Epoch 25/150
856/856 [==============================] - 0s 154us/sample - loss: 1.0021 - acc: 0.5748 - val_loss: 0.8782 - val_acc: 0.6465
Epoch 26/150
856/856 [==============================] - 0s 139us/sample - loss: 1.0041 - acc: 0.5864 - val_loss: 0.8828 - val_acc: 0.6651
Epoch 27/150
856/856 [==============================] - 0s 130us/sample - loss: 1.0035 - acc: 0.5701 - val_loss: 0.8859 - val_acc: 0.6372
Epoch 28/150
856/856 [==============================] - 0s 129us/sample - loss: 1.0044 - acc: 0.5689 - val_loss: 0.8832 - val_acc: 0.6465
Epoch 29/150
856/856 [==============================] - 0s 125us/sample - loss: 0.9966 - acc: 0.5794 - val_loss: 0.8864 - val_acc: 0.6651
Epoch 30/150
856/856 [==============================] - 0s 162us/sample - loss: 0.9984 - acc: 0.5829 - val_loss: 0.8831 - val_acc: 0.6512
Epoch 31/150
856/856 [==============================] - 0s 160us/sample - loss: 0.9877 - acc: 0.5818 - val_loss: 0.8798 - val_acc: 0.6605
Epoch 32/150
856/856 [==============================] - 0s 132us/sample - loss: 1.0035 - acc: 0.5771 - val_loss: 0.8785 - val_acc: 0.6465
Epoch 33/150
856/856 [==============================] - 0s 127us/sample - loss: 0.9959 - acc: 0.5724 - val_loss: 0.8823 - val_acc: 0.6558
Epoch 34/150
856/856 [==============================] - 0s 122us/sample - loss: 1.0147 - acc: 0.5724 - val_loss: 0.8821 - val_acc: 0.6558
Epoch 35/150
856/856 [==============================] - 0s 139us/sample - loss: 1.0020 - acc: 0.5876 - val_loss: 0.8830 - val_acc: 0.6744
Epoch 36/150
856/856 [==============================] - 0s 167us/sample - loss: 0.9955 - acc: 0.5864 - val_loss: 0.8791 - val_acc: 0.6744
Epoch 37/150
856/856 [==============================] - 0s 126us/sample - loss: 0.9920 - acc: 0.5794 - val_loss: 0.8703 - val_acc: 0.6651
Epoch 38/150
856/856 [==============================] - 0s 132us/sample - loss: 1.0089 - acc: 0.5841 - val_loss: 0.8747 - val_acc: 0.6698
Epoch 39/150
856/856 [==============================] - 0s 145us/sample - loss: 1.0003 - acc: 0.5900 - val_loss: 0.8795 - val_acc: 0.6558
Epoch 40/150
856/856 [==============================] - 0s 129us/sample - loss: 0.9956 - acc: 0.5958 - val_loss: 0.8771 - val_acc: 0.6605
Epoch 41/150
856/856 [==============================] - 0s 119us/sample - loss: 0.9826 - acc: 0.6098 - val_loss: 0.8790 - val_acc: 0.6558
Epoch 42/150
856/856 [==============================] - 0s 125us/sample - loss: 0.9747 - acc: 0.6040 - val_loss: 0.8682 - val_acc: 0.6512
Epoch 43/150
856/856 [==============================] - 0s 127us/sample - loss: 0.9917 - acc: 0.5724 - val_loss: 0.8712 - val_acc: 0.6698
Epoch 44/150
856/856 [==============================] - 0s 125us/sample - loss: 0.9865 - acc: 0.5783 - val_loss: 0.8773 - val_acc: 0.6512
Epoch 45/150
856/856 [==============================] - 0s 136us/sample - loss: 0.9749 - acc: 0.5946 - val_loss: 0.8713 - val_acc: 0.6419
Epoch 46/150
856/856 [==============================] - 0s 138us/sample - loss: 0.9869 - acc: 0.5829 - val_loss: 0.8770 - val_acc: 0.6512
Epoch 47/150
856/856 [==============================] - 0s 133us/sample - loss: 0.9855 - acc: 0.6086 - val_loss: 0.8802 - val_acc: 0.6512
Epoch 48/150
856/856 [==============================] - 0s 126us/sample - loss: 0.9728 - acc: 0.6157 - val_loss: 0.8731 - val_acc: 0.6605
Epoch 49/150
856/856 [==============================] - 0s 127us/sample - loss: 0.9993 - acc: 0.5853 - val_loss: 0.8827 - val_acc: 0.6372
Epoch 50/150
856/856 [==============================] - 0s 127us/sample - loss: 0.9820 - acc: 0.5829 - val_loss: 0.8754 - val_acc: 0.6465
Epoch 51/150
856/856 [==============================] - 0s 161us/sample - loss: 0.9809 - acc: 0.5818 - val_loss: 0.8767 - val_acc: 0.6465
Epoch 52/150
856/856 [==============================] - 0s 153us/sample - loss: 0.9749 - acc: 0.5841 - val_loss: 0.8717 - val_acc: 0.6605
Epoch 53/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9896 - acc: 0.5829 - val_loss: 0.8788 - val_acc: 0.6326
Epoch 54/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9595 - acc: 0.5958 - val_loss: 0.8742 - val_acc: 0.6465
Epoch 55/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9602 - acc: 0.5970 - val_loss: 0.8718 - val_acc: 0.6605
Epoch 56/150
856/856 [==============================] - 0s 110us/sample - loss: 0.9819 - acc: 0.5829 - val_loss: 0.8699 - val_acc: 0.6512
Epoch 57/150
856/856 [==============================] - 0s 111us/sample - loss: 0.9743 - acc: 0.5970 - val_loss: 0.8721 - val_acc: 0.6512
Epoch 58/150
856/856 [==============================] - 0s 116us/sample - loss: 0.9678 - acc: 0.5911 - val_loss: 0.8646 - val_acc: 0.6558
Epoch 59/150
856/856 [==============================] - 0s 123us/sample - loss: 0.9820 - acc: 0.5958 - val_loss: 0.8675 - val_acc: 0.6605
Epoch 60/150
856/856 [==============================] - 0s 114us/sample - loss: 0.9694 - acc: 0.5888 - val_loss: 0.8735 - val_acc: 0.6465
Epoch 61/150
856/856 [==============================] - 0s 114us/sample - loss: 0.9825 - acc: 0.5771 - val_loss: 0.8709 - val_acc: 0.6558
Epoch 62/150
856/856 [==============================] - 0s 127us/sample - loss: 0.9681 - acc: 0.5888 - val_loss: 0.8697 - val_acc: 0.6512
Epoch 63/150
856/856 [==============================] - 0s 140us/sample - loss: 0.9670 - acc: 0.5923 - val_loss: 0.8739 - val_acc: 0.6326
Epoch 64/150
856/856 [==============================] - 0s 116us/sample - loss: 0.9857 - acc: 0.5900 - val_loss: 0.8758 - val_acc: 0.6605
Epoch 65/150
856/856 [==============================] - 0s 116us/sample - loss: 0.9751 - acc: 0.5911 - val_loss: 0.8686 - val_acc: 0.6605
Epoch 66/150
856/856 [==============================] - 0s 118us/sample - loss: 0.9737 - acc: 0.6040 - val_loss: 0.8712 - val_acc: 0.6651
Epoch 67/150
856/856 [==============================] - 0s 112us/sample - loss: 0.9617 - acc: 0.6016 - val_loss: 0.8722 - val_acc: 0.6651
Epoch 68/150
856/856 [==============================] - 0s 123us/sample - loss: 0.9901 - acc: 0.5888 - val_loss: 0.8727 - val_acc: 0.6698
Epoch 69/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9455 - acc: 0.5958 - val_loss: 0.8702 - val_acc: 0.6512
Epoch 70/150
856/856 [==============================] - 0s 126us/sample - loss: 0.9436 - acc: 0.5993 - val_loss: 0.8644 - val_acc: 0.6605
Epoch 71/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9690 - acc: 0.6005 - val_loss: 0.8692 - val_acc: 0.6605
Epoch 72/150
856/856 [==============================] - 0s 112us/sample - loss: 0.9507 - acc: 0.6215 - val_loss: 0.8663 - val_acc: 0.6651
Epoch 73/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9816 - acc: 0.6028 - val_loss: 0.8687 - val_acc: 0.6558
Epoch 74/150
856/856 [==============================] - 0s 168us/sample - loss: 0.9722 - acc: 0.5841 - val_loss: 0.8696 - val_acc: 0.6698
Epoch 75/150
856/856 [==============================] - 0s 123us/sample - loss: 0.9752 - acc: 0.5853 - val_loss: 0.8747 - val_acc: 0.6419
Epoch 76/150
856/856 [==============================] - 0s 186us/sample - loss: 0.9656 - acc: 0.5900 - val_loss: 0.8708 - val_acc: 0.6558
Epoch 77/150
856/856 [==============================] - 0s 151us/sample - loss: 0.9600 - acc: 0.6121 - val_loss: 0.8715 - val_acc: 0.6605
Epoch 78/150
856/856 [==============================] - 0s 127us/sample - loss: 0.9562 - acc: 0.5935 - val_loss: 0.8730 - val_acc: 0.6465
Epoch 79/150
856/856 [==============================] - 0s 130us/sample - loss: 0.9792 - acc: 0.5783 - val_loss: 0.8710 - val_acc: 0.6465
Epoch 80/150
856/856 [==============================] - 0s 174us/sample - loss: 0.9624 - acc: 0.5981 - val_loss: 0.8685 - val_acc: 0.6512
Epoch 81/150
856/856 [==============================] - 0s 124us/sample - loss: 0.9531 - acc: 0.6145 - val_loss: 0.8728 - val_acc: 0.6512
Epoch 82/150
856/856 [==============================] - 0s 112us/sample - loss: 0.9591 - acc: 0.5900 - val_loss: 0.8737 - val_acc: 0.6465
Epoch 83/150
856/856 [==============================] - 0s 111us/sample - loss: 0.9731 - acc: 0.5981 - val_loss: 0.8842 - val_acc: 0.6512
Epoch 84/150
856/856 [==============================] - 0s 114us/sample - loss: 0.9606 - acc: 0.5888 - val_loss: 0.8727 - val_acc: 0.6465
Epoch 85/150
856/856 [==============================] - 0s 112us/sample - loss: 0.9357 - acc: 0.6005 - val_loss: 0.8658 - val_acc: 0.6465
Epoch 86/150
856/856 [==============================] - 0s 132us/sample - loss: 0.9690 - acc: 0.6040 - val_loss: 0.8674 - val_acc: 0.6512
Epoch 87/150
856/856 [==============================] - 0s 131us/sample - loss: 0.9645 - acc: 0.5958 - val_loss: 0.8681 - val_acc: 0.6605
Epoch 88/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9596 - acc: 0.5864 - val_loss: 0.8719 - val_acc: 0.6605
Epoch 89/150
856/856 [==============================] - 0s 118us/sample - loss: 0.9593 - acc: 0.6110 - val_loss: 0.8715 - val_acc: 0.6512
Epoch 90/150
856/856 [==============================] - 0s 122us/sample - loss: 0.9555 - acc: 0.6110 - val_loss: 0.8717 - val_acc: 0.6558
Epoch 91/150
856/856 [==============================] - 0s 146us/sample - loss: 0.9570 - acc: 0.6075 - val_loss: 0.8722 - val_acc: 0.6419
Epoch 92/150
856/856 [==============================] - 0s 130us/sample - loss: 0.9488 - acc: 0.6028 - val_loss: 0.8730 - val_acc: 0.6419
Epoch 93/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9499 - acc: 0.5970 - val_loss: 0.8615 - val_acc: 0.6419
Epoch 94/150
856/856 [==============================] - 0s 113us/sample - loss: 0.9408 - acc: 0.6098 - val_loss: 0.8602 - val_acc: 0.6558
Epoch 95/150
856/856 [==============================] - 0s 132us/sample - loss: 0.9630 - acc: 0.6051 - val_loss: 0.8668 - val_acc: 0.6512
Epoch 96/150
856/856 [==============================] - 0s 125us/sample - loss: 0.9642 - acc: 0.6121 - val_loss: 0.8704 - val_acc: 0.6558
Epoch 97/150
856/856 [==============================] - 0s 200us/sample - loss: 0.9589 - acc: 0.5935 - val_loss: 0.8699 - val_acc: 0.6512
Epoch 98/150
856/856 [==============================] - 0s 132us/sample - loss: 0.9424 - acc: 0.6005 - val_loss: 0.8649 - val_acc: 0.6558
Epoch 99/150
856/856 [==============================] - 0s 113us/sample - loss: 0.9492 - acc: 0.6005 - val_loss: 0.8702 - val_acc: 0.6558
Epoch 100/150
856/856 [==============================] - 0s 130us/sample - loss: 0.9386 - acc: 0.6168 - val_loss: 0.8691 - val_acc: 0.6558
Epoch 101/150
856/856 [==============================] - 0s 155us/sample - loss: 0.9396 - acc: 0.6238 - val_loss: 0.8610 - val_acc: 0.6605
Epoch 102/150
856/856 [==============================] - 0s 124us/sample - loss: 0.9517 - acc: 0.6133 - val_loss: 0.8652 - val_acc: 0.6558
Epoch 103/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9519 - acc: 0.6016 - val_loss: 0.8693 - val_acc: 0.6512
Epoch 104/150
856/856 [==============================] - 0s 138us/sample - loss: 0.9498 - acc: 0.6098 - val_loss: 0.8775 - val_acc: 0.6326
Epoch 105/150
856/856 [==============================] - 0s 140us/sample - loss: 0.9649 - acc: 0.6040 - val_loss: 0.8702 - val_acc: 0.6605
Epoch 106/150
856/856 [==============================] - 0s 136us/sample - loss: 0.9450 - acc: 0.6086 - val_loss: 0.8684 - val_acc: 0.6512
Epoch 107/150
856/856 [==============================] - 0s 138us/sample - loss: 0.9541 - acc: 0.6075 - val_loss: 0.8670 - val_acc: 0.6698
Epoch 108/150
856/856 [==============================] - 0s 129us/sample - loss: 0.9534 - acc: 0.6145 - val_loss: 0.8619 - val_acc: 0.6512
Epoch 109/150
856/856 [==============================] - 0s 113us/sample - loss: 0.9490 - acc: 0.5958 - val_loss: 0.8650 - val_acc: 0.6512
Epoch 110/150
856/856 [==============================] - 0s 152us/sample - loss: 0.9353 - acc: 0.5970 - val_loss: 0.8640 - val_acc: 0.6651
Epoch 111/150
856/856 [==============================] - 0s 161us/sample - loss: 0.9511 - acc: 0.6285 - val_loss: 0.8672 - val_acc: 0.6419
Epoch 112/150
856/856 [==============================] - 0s 117us/sample - loss: 0.9558 - acc: 0.6040 - val_loss: 0.8672 - val_acc: 0.6512
Epoch 113/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9415 - acc: 0.5970 - val_loss: 0.8686 - val_acc: 0.6465
Epoch 114/150
856/856 [==============================] - 0s 117us/sample - loss: 0.9366 - acc: 0.5981 - val_loss: 0.8691 - val_acc: 0.6279
Epoch 115/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9267 - acc: 0.6238 - val_loss: 0.8629 - val_acc: 0.6465
Epoch 116/150
856/856 [==============================] - 0s 114us/sample - loss: 0.9373 - acc: 0.6215 - val_loss: 0.8618 - val_acc: 0.6558
Epoch 117/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9534 - acc: 0.6016 - val_loss: 0.8622 - val_acc: 0.6558
Epoch 118/150
856/856 [==============================] - 0s 131us/sample - loss: 0.9410 - acc: 0.5935 - val_loss: 0.8637 - val_acc: 0.6558
Epoch 119/150
856/856 [==============================] - 0s 140us/sample - loss: 0.9423 - acc: 0.6075 - val_loss: 0.8641 - val_acc: 0.6512
Epoch 120/150
856/856 [==============================] - 0s 130us/sample - loss: 0.9346 - acc: 0.6075 - val_loss: 0.8653 - val_acc: 0.6512
Epoch 121/150
856/856 [==============================] - 0s 138us/sample - loss: 0.9573 - acc: 0.5981 - val_loss: 0.8591 - val_acc: 0.6465
Epoch 122/150
856/856 [==============================] - 0s 140us/sample - loss: 0.9438 - acc: 0.6098 - val_loss: 0.8592 - val_acc: 0.6651
Epoch 123/150
856/856 [==============================] - 0s 117us/sample - loss: 0.9487 - acc: 0.6098 - val_loss: 0.8662 - val_acc: 0.6465
Epoch 124/150
856/856 [==============================] - 0s 116us/sample - loss: 0.9483 - acc: 0.5911 - val_loss: 0.8639 - val_acc: 0.6558
Epoch 125/150
856/856 [==============================] - 0s 125us/sample - loss: 0.9169 - acc: 0.6238 - val_loss: 0.8597 - val_acc: 0.6512
Epoch 126/150
856/856 [==============================] - 0s 155us/sample - loss: 0.9504 - acc: 0.5946 - val_loss: 0.8632 - val_acc: 0.6419
Epoch 127/150
856/856 [==============================] - 0s 131us/sample - loss: 0.9528 - acc: 0.6157 - val_loss: 0.8648 - val_acc: 0.6558
Epoch 128/150
856/856 [==============================] - 0s 118us/sample - loss: 0.9314 - acc: 0.6203 - val_loss: 0.8619 - val_acc: 0.6419
Epoch 129/150
856/856 [==============================] - 0s 116us/sample - loss: 0.9409 - acc: 0.6063 - val_loss: 0.8705 - val_acc: 0.6512
Epoch 130/150
856/856 [==============================] - 0s 113us/sample - loss: 0.9437 - acc: 0.6110 - val_loss: 0.8646 - val_acc: 0.6651
Epoch 131/150
856/856 [==============================] - 0s 122us/sample - loss: 0.9211 - acc: 0.6110 - val_loss: 0.8608 - val_acc: 0.6558
Epoch 132/150
856/856 [==============================] - 0s 119us/sample - loss: 0.9344 - acc: 0.6040 - val_loss: 0.8660 - val_acc: 0.6512
Epoch 133/150
856/856 [==============================] - 0s 129us/sample - loss: 0.9382 - acc: 0.6005 - val_loss: 0.8674 - val_acc: 0.6558
Epoch 134/150
856/856 [==============================] - 0s 151us/sample - loss: 0.9451 - acc: 0.5923 - val_loss: 0.8673 - val_acc: 0.6465
Epoch 135/150
856/856 [==============================] - 0s 127us/sample - loss: 0.9412 - acc: 0.6192 - val_loss: 0.8658 - val_acc: 0.6419
Epoch 136/150
856/856 [==============================] - 0s 126us/sample - loss: 0.9436 - acc: 0.6215 - val_loss: 0.8679 - val_acc: 0.6465
Epoch 137/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9433 - acc: 0.6110 - val_loss: 0.8732 - val_acc: 0.6512
Epoch 138/150
856/856 [==============================] - 0s 157us/sample - loss: 0.9337 - acc: 0.6063 - val_loss: 0.8646 - val_acc: 0.6465
Epoch 139/150
856/856 [==============================] - 0s 124us/sample - loss: 0.9507 - acc: 0.6016 - val_loss: 0.8694 - val_acc: 0.6605
Epoch 140/150
856/856 [==============================] - 0s 122us/sample - loss: 0.9213 - acc: 0.6273 - val_loss: 0.8667 - val_acc: 0.6558
Epoch 141/150
856/856 [==============================] - 0s 123us/sample - loss: 0.9226 - acc: 0.6121 - val_loss: 0.8718 - val_acc: 0.6512
Epoch 142/150
856/856 [==============================] - 0s 218us/sample - loss: 0.9379 - acc: 0.5935 - val_loss: 0.8704 - val_acc: 0.6558
Epoch 143/150
856/856 [==============================] - 0s 120us/sample - loss: 0.9350 - acc: 0.6051 - val_loss: 0.8728 - val_acc: 0.6512
Epoch 144/150
856/856 [==============================] - 0s 113us/sample - loss: 0.9227 - acc: 0.6250 - val_loss: 0.8671 - val_acc: 0.6651
Epoch 145/150
856/856 [==============================] - 0s 111us/sample - loss: 0.9487 - acc: 0.6157 - val_loss: 0.8681 - val_acc: 0.6512
Epoch 146/150
856/856 [==============================] - 0s 123us/sample - loss: 0.9183 - acc: 0.6168 - val_loss: 0.8679 - val_acc: 0.6465
Epoch 147/150
856/856 [==============================] - 0s 118us/sample - loss: 0.9373 - acc: 0.6121 - val_loss: 0.8706 - val_acc: 0.6465
Epoch 148/150
856/856 [==============================] - 0s 113us/sample - loss: 0.9354 - acc: 0.6110 - val_loss: 0.8701 - val_acc: 0.6419
Epoch 149/150
856/856 [==============================] - 0s 116us/sample - loss: 0.9287 - acc: 0.6297 - val_loss: 0.8675 - val_acc: 0.6512
Epoch 150/150
856/856 [==============================] - 0s 161us/sample - loss: 0.9290 - acc: 0.6040 - val_loss: 0.8718 - val_acc: 0.6465
    In [137]:
    print(history.history.keys())
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.show()

    dict_keys(['loss', 'acc', 'val_loss', 'val_acc'])
    In [139]:
    #y_pred = model.predict(X_test)
y_pred = model.predict_classes(X_test)
    In [140]:
    y_classes = [np.argmax(y, axis=None, out=None) for y in y_test]
    In [144]:
    y_train_classes = [np.argmax(y, axis=None, out=None) for y in y_train]

    Train features count

    In [150]:
    sns.countplot(y_train_classes)
    Out[150]:
    <matplotlib.axes._subplots.AxesSubplot at 0x66f2ca08>

    Test Features count

    In [151]:
    sns.countplot(y_classes)
    Out[151]:
    <matplotlib.axes._subplots.AxesSubplot at 0x670d0f08>

    Prediction feature count

    In [152]:
    sns.countplot(y_pred)
    Out[152]:
    <matplotlib.axes._subplots.AxesSubplot at 0x69622388>
    In [141]:
    from sklearn.metrics import classification_report
print(classification_report(y_pred,y_classes))

                  precision    recall  f1-score   support

           3       0.00      0.00      0.00         0
           4       0.00      0.00      0.00         0
           5       0.75      0.62      0.68       262
           6       0.46      0.52      0.49       189
           7       0.53      0.48      0.50        77
           8       0.00      0.00      0.00         0

    accuracy                           0.56       528
   macro avg       0.29      0.27      0.28       528
weighted avg       0.61      0.56      0.59       528

    Observations:

  • Since, the data is heavily imbalanced we will use precion and recall scores to judge the predicatability of our model
  • The model trained on PCA can predict classes 5,6 and 7 with recall score ranging from 0.50 to 0.62 but it is not able to predict classes 3,4 and 8(which have lower count in the dataset) accurately. This brings the weighted average down to 0.56.

    • Training neural net on original dataset

    In [266]:
    for column in df1.drop(columns=['Signal_Strength']).columns:
    df1[column] = df1[column].apply(lambda x:
                        (x - df1[column].mean())/df1[column].std()
                       )
    In [267]:
    categorical_labels = to_categorical(df1['Signal_Strength'], num_classes=None)
    In [281]:
    X = df1.drop(columns=['Signal_Strength']).values
Y = categorical_labels
X_train, X_test, y_train, y_test = train_test_split(X, Y,test_size=0.33, random_state=42)
    In [282]:
    y_train_classes = [np.argmax(y, axis=None, out=None) for y in y_train]
    In [283]:
    from imblearn.over_sampling import SMOTE
sm = SMOTE(random_state=2)
X_train_res,y_train_res = sm.fit_sample(X_train,y_train_classes)
    In [284]:
    y_train_res = to_categorical(y_train_res,num_classes=None)
    In [285]:
    y_train_classes = [np.argmax(y, axis=None, out=None) for y in y_train_res]
    In [286]:
    model1 = Sequential()
model1.add(Dense(20, input_dim=11, kernel_initializer='normal', activation='relu'))
model1.add(Dropout(0.25))
model1.add(Dense(20, activation='relu'))
model1.add(Dropout(0.25))
model1.add(Dense(20, activation='relu'))
model1.add(Dropout(0.25))
model1.add(Dense(9, activation='softmax'))
model1.summary()

    Model: "sequential_23"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_77 (Dense)             (None, 20)                240       
_________________________________________________________________
dropout_54 (Dropout)         (None, 20)                0         
_________________________________________________________________
dense_78 (Dense)             (None, 20)                420       
_________________________________________________________________
dropout_55 (Dropout)         (None, 20)                0         
_________________________________________________________________
dense_79 (Dense)             (None, 20)                420       
_________________________________________________________________
dropout_56 (Dropout)         (None, 20)                0         
_________________________________________________________________
dense_80 (Dense)             (None, 9)                 189       
=================================================================
Total params: 1,269
Trainable params: 1,269
Non-trainable params: 0
_________________________________________________________________
    In [287]:
    from tensorflow import keras
lr_schedule = keras.optimizers.schedules.ExponentialDecay(
    initial_learning_rate=1e-2,
    decay_steps=10000,
    decay_rate=0.9)
sgd = keras.optimizers.SGD(learning_rate=lr_schedule)
model1.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
    In [288]:
    history = model1.fit(X_train_res, y_train_res, epochs=200, batch_size=10,  verbose=1, validation_split=0.2)

    Train on 2227 samples, validate on 557 samples
Epoch 1/200
2227/2227 [==============================] - 1s 393us/sample - loss: 2.0198 - acc: 0.2326 - val_loss: 2.5098 - val_acc: 0.1400
Epoch 2/200
2227/2227 [==============================] - 0s 154us/sample - loss: 1.7571 - acc: 0.3035 - val_loss: 3.0194 - val_acc: 0.0952
Epoch 3/200
2227/2227 [==============================] - 0s 161us/sample - loss: 1.5943 - acc: 0.3408 - val_loss: 3.5634 - val_acc: 0.0862
Epoch 4/200
2227/2227 [==============================] - 0s 156us/sample - loss: 1.4632 - acc: 0.3754 - val_loss: 4.1195 - val_acc: 0.0305
Epoch 5/200
2227/2227 [==============================] - 0s 154us/sample - loss: 1.3965 - acc: 0.4109 - val_loss: 4.4781 - val_acc: 0.1364
Epoch 6/200
2227/2227 [==============================] - 0s 154us/sample - loss: 1.3567 - acc: 0.4284 - val_loss: 4.5553 - val_acc: 0.1131
Epoch 7/200
2227/2227 [==============================] - 0s 159us/sample - loss: 1.3271 - acc: 0.4387 - val_loss: 4.7686 - val_acc: 0.1257
Epoch 8/200
2227/2227 [==============================] - 0s 177us/sample - loss: 1.2737 - acc: 0.4656 - val_loss: 4.9856 - val_acc: 0.1167
Epoch 9/200
2227/2227 [==============================] - 0s 167us/sample - loss: 1.2620 - acc: 0.4576 - val_loss: 5.0313 - val_acc: 0.1364
Epoch 10/200
2227/2227 [==============================] - 0s 163us/sample - loss: 1.2287 - acc: 0.4737 - val_loss: 5.0977 - val_acc: 0.1436
Epoch 11/200
2227/2227 [==============================] - 0s 178us/sample - loss: 1.2251 - acc: 0.4877 - val_loss: 5.1169 - val_acc: 0.1149
Epoch 12/200
2227/2227 [==============================] - 0s 172us/sample - loss: 1.2174 - acc: 0.4755 - val_loss: 4.9095 - val_acc: 0.1005
Epoch 13/200
2227/2227 [==============================] - 0s 158us/sample - loss: 1.2097 - acc: 0.4841 - val_loss: 4.6942 - val_acc: 0.1149
Epoch 14/200
2227/2227 [==============================] - 0s 178us/sample - loss: 1.2011 - acc: 0.4953 - val_loss: 4.7584 - val_acc: 0.1382
Epoch 15/200
2227/2227 [==============================] - 0s 172us/sample - loss: 1.1904 - acc: 0.4966 - val_loss: 4.5348 - val_acc: 0.1203
Epoch 16/200
2227/2227 [==============================] - 0s 153us/sample - loss: 1.1666 - acc: 0.5025 - val_loss: 4.4723 - val_acc: 0.1167
Epoch 17/200
2227/2227 [==============================] - 0s 161us/sample - loss: 1.1542 - acc: 0.5177 - val_loss: 4.4530 - val_acc: 0.0969
Epoch 18/200
2227/2227 [==============================] - 0s 181us/sample - loss: 1.1458 - acc: 0.5173 - val_loss: 4.4160 - val_acc: 0.1239
Epoch 19/200
2227/2227 [==============================] - 0s 165us/sample - loss: 1.1419 - acc: 0.5209 - val_loss: 4.3840 - val_acc: 0.0539
Epoch 20/200
2227/2227 [==============================] - 0s 153us/sample - loss: 1.1439 - acc: 0.5106 - val_loss: 4.2462 - val_acc: 0.1329
Epoch 21/200
2227/2227 [==============================] - 0s 188us/sample - loss: 1.1344 - acc: 0.5236 - val_loss: 4.0229 - val_acc: 0.1203
Epoch 22/200
2227/2227 [==============================] - 0s 155us/sample - loss: 1.1117 - acc: 0.5519 - val_loss: 4.0518 - val_acc: 0.0987
Epoch 23/200
2227/2227 [==============================] - 0s 167us/sample - loss: 1.1021 - acc: 0.5285 - val_loss: 4.1029 - val_acc: 0.1257
Epoch 24/200
2227/2227 [==============================] - 0s 180us/sample - loss: 1.1074 - acc: 0.5393 - val_loss: 3.9681 - val_acc: 0.1005
Epoch 25/200
2227/2227 [==============================] - 0s 185us/sample - loss: 1.0908 - acc: 0.5442 - val_loss: 3.8953 - val_acc: 0.1311
Epoch 26/200
2227/2227 [==============================] - 0s 205us/sample - loss: 1.0908 - acc: 0.5384 - val_loss: 3.9189 - val_acc: 0.1131
Epoch 27/200
2227/2227 [==============================] - 0s 163us/sample - loss: 1.0895 - acc: 0.5465 - val_loss: 3.8725 - val_acc: 0.0880
Epoch 28/200
2227/2227 [==============================] - 0s 159us/sample - loss: 1.1105 - acc: 0.5415 - val_loss: 3.7097 - val_acc: 0.1203
Epoch 29/200
2227/2227 [==============================] - 0s 154us/sample - loss: 1.0705 - acc: 0.5550 - val_loss: 3.8420 - val_acc: 0.1329
Epoch 30/200
2227/2227 [==============================] - 0s 158us/sample - loss: 1.0690 - acc: 0.5528 - val_loss: 3.7532 - val_acc: 0.1185
Epoch 31/200
2227/2227 [==============================] - 0s 158us/sample - loss: 1.0459 - acc: 0.5510 - val_loss: 3.5311 - val_acc: 0.1239
Epoch 32/200
2227/2227 [==============================] - 0s 158us/sample - loss: 1.0632 - acc: 0.5608 - val_loss: 3.6312 - val_acc: 0.1185
Epoch 33/200
2227/2227 [==============================] - 0s 165us/sample - loss: 1.0537 - acc: 0.5577 - val_loss: 3.6048 - val_acc: 0.1149
Epoch 34/200
2227/2227 [==============================] - 0s 199us/sample - loss: 1.0656 - acc: 0.5604 - val_loss: 3.6391 - val_acc: 0.1257
Epoch 35/200
2227/2227 [==============================] - 0s 194us/sample - loss: 1.0508 - acc: 0.5644 - val_loss: 3.6019 - val_acc: 0.1239
Epoch 36/200
2227/2227 [==============================] - 0s 159us/sample - loss: 1.0360 - acc: 0.5667 - val_loss: 3.5419 - val_acc: 0.1203
Epoch 37/200
2227/2227 [==============================] - 0s 156us/sample - loss: 1.0261 - acc: 0.5626 - val_loss: 3.5454 - val_acc: 0.1167
Epoch 38/200
2227/2227 [==============================] - 0s 166us/sample - loss: 1.0434 - acc: 0.5617 - val_loss: 3.4077 - val_acc: 0.1275
Epoch 39/200
2227/2227 [==============================] - 0s 172us/sample - loss: 1.0130 - acc: 0.5779 - val_loss: 3.4967 - val_acc: 0.1131
Epoch 40/200
2227/2227 [==============================] - 0s 185us/sample - loss: 1.0233 - acc: 0.5757 - val_loss: 3.4735 - val_acc: 0.1095
Epoch 41/200
2227/2227 [==============================] - 0s 192us/sample - loss: 1.0138 - acc: 0.5824 - val_loss: 3.4408 - val_acc: 0.0916
Epoch 42/200
2227/2227 [==============================] - 0s 173us/sample - loss: 1.0139 - acc: 0.5680 - val_loss: 3.4815 - val_acc: 0.1059
Epoch 43/200
2227/2227 [==============================] - 0s 173us/sample - loss: 1.0163 - acc: 0.5828 - val_loss: 3.3293 - val_acc: 0.1203
Epoch 44/200
2227/2227 [==============================] - 0s 176us/sample - loss: 0.9944 - acc: 0.5815 - val_loss: 3.2040 - val_acc: 0.0844
Epoch 45/200
2227/2227 [==============================] - 0s 167us/sample - loss: 0.9983 - acc: 0.5698 - val_loss: 3.4153 - val_acc: 0.1167
Epoch 46/200
2227/2227 [==============================] - 0s 193us/sample - loss: 1.0038 - acc: 0.5802 - val_loss: 3.5126 - val_acc: 0.1185
Epoch 47/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.9993 - acc: 0.5739 - val_loss: 3.5727 - val_acc: 0.1113
Epoch 48/200
2227/2227 [==============================] - 0s 187us/sample - loss: 1.0013 - acc: 0.5828 - val_loss: 3.5892 - val_acc: 0.1203
Epoch 49/200
2227/2227 [==============================] - 0s 163us/sample - loss: 0.9540 - acc: 0.6022 - val_loss: 3.7001 - val_acc: 0.1203
Epoch 50/200
2227/2227 [==============================] - 0s 158us/sample - loss: 0.9870 - acc: 0.5963 - val_loss: 3.5774 - val_acc: 0.0557
Epoch 51/200
2227/2227 [==============================] - 0s 163us/sample - loss: 0.9712 - acc: 0.5846 - val_loss: 3.5321 - val_acc: 0.0969
Epoch 52/200
2227/2227 [==============================] - 0s 155us/sample - loss: 0.9888 - acc: 0.5914 - val_loss: 3.3612 - val_acc: 0.0952
Epoch 53/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.9617 - acc: 0.5954 - val_loss: 3.3292 - val_acc: 0.0880
Epoch 54/200
2227/2227 [==============================] - 0s 157us/sample - loss: 0.9648 - acc: 0.5968 - val_loss: 3.4007 - val_acc: 0.1167
Epoch 55/200
2227/2227 [==============================] - 0s 148us/sample - loss: 0.9517 - acc: 0.6026 - val_loss: 3.4119 - val_acc: 0.1095
Epoch 56/200
2227/2227 [==============================] - 0s 190us/sample - loss: 0.9431 - acc: 0.6035 - val_loss: 3.4207 - val_acc: 0.1113
Epoch 57/200
2227/2227 [==============================] - 0s 181us/sample - loss: 0.9679 - acc: 0.5995 - val_loss: 3.4373 - val_acc: 0.1113
Epoch 58/200
2227/2227 [==============================] - 0s 173us/sample - loss: 0.9734 - acc: 0.6035 - val_loss: 3.3675 - val_acc: 0.0952
Epoch 59/200
2227/2227 [==============================] - 0s 180us/sample - loss: 0.9456 - acc: 0.6062 - val_loss: 3.4285 - val_acc: 0.1113
Epoch 60/200
2227/2227 [==============================] - 0s 176us/sample - loss: 0.9408 - acc: 0.6125 - val_loss: 3.4801 - val_acc: 0.1221
Epoch 61/200
2227/2227 [==============================] - 0s 190us/sample - loss: 0.9482 - acc: 0.6004 - val_loss: 3.3930 - val_acc: 0.0862
Epoch 62/200
2227/2227 [==============================] - 0s 167us/sample - loss: 0.9410 - acc: 0.5999 - val_loss: 3.2920 - val_acc: 0.1023
Epoch 63/200
2227/2227 [==============================] - 0s 172us/sample - loss: 0.9443 - acc: 0.6031 - val_loss: 3.3909 - val_acc: 0.1149
Epoch 64/200
2227/2227 [==============================] - 0s 186us/sample - loss: 0.9182 - acc: 0.6102 - val_loss: 3.4333 - val_acc: 0.1095
Epoch 65/200
2227/2227 [==============================] - 0s 196us/sample - loss: 0.9334 - acc: 0.6062 - val_loss: 3.4077 - val_acc: 0.1167
Epoch 66/200
2227/2227 [==============================] - 1s 234us/sample - loss: 0.9251 - acc: 0.6075 - val_loss: 3.4393 - val_acc: 0.1203
Epoch 67/200
2227/2227 [==============================] - 0s 175us/sample - loss: 0.9213 - acc: 0.6170 - val_loss: 3.4204 - val_acc: 0.1149
Epoch 68/200
2227/2227 [==============================] - 0s 165us/sample - loss: 0.9229 - acc: 0.6089 - val_loss: 3.2635 - val_acc: 0.0898
Epoch 69/200
2227/2227 [==============================] - 0s 172us/sample - loss: 0.9226 - acc: 0.6111 - val_loss: 3.3604 - val_acc: 0.1185
Epoch 70/200
2227/2227 [==============================] - 0s 170us/sample - loss: 0.9172 - acc: 0.6044 - val_loss: 3.3547 - val_acc: 0.1257
Epoch 71/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.9252 - acc: 0.6170 - val_loss: 3.3088 - val_acc: 0.1131
Epoch 72/200
2227/2227 [==============================] - 0s 170us/sample - loss: 0.9058 - acc: 0.6273 - val_loss: 3.2979 - val_acc: 0.1239
Epoch 73/200
2227/2227 [==============================] - 0s 166us/sample - loss: 0.8998 - acc: 0.6264 - val_loss: 3.2219 - val_acc: 0.1239
Epoch 74/200
2227/2227 [==============================] - 0s 180us/sample - loss: 0.9185 - acc: 0.6107 - val_loss: 3.1362 - val_acc: 0.1095
Epoch 75/200
2227/2227 [==============================] - 0s 181us/sample - loss: 0.8987 - acc: 0.6318 - val_loss: 3.1706 - val_acc: 0.1167
Epoch 76/200
2227/2227 [==============================] - 0s 159us/sample - loss: 0.9031 - acc: 0.6206 - val_loss: 3.2503 - val_acc: 0.1131
Epoch 77/200
2227/2227 [==============================] - 0s 170us/sample - loss: 0.8877 - acc: 0.6228 - val_loss: 3.3912 - val_acc: 0.1203
Epoch 78/200
2227/2227 [==============================] - 0s 171us/sample - loss: 0.9031 - acc: 0.6291 - val_loss: 3.2938 - val_acc: 0.1167
Epoch 79/200
2227/2227 [==============================] - 0s 159us/sample - loss: 0.8959 - acc: 0.6201 - val_loss: 3.3708 - val_acc: 0.1221
Epoch 80/200
2227/2227 [==============================] - 0s 157us/sample - loss: 0.8902 - acc: 0.6309 - val_loss: 3.2409 - val_acc: 0.1203
Epoch 81/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8633 - acc: 0.6322 - val_loss: 3.2070 - val_acc: 0.1364
Epoch 82/200
2227/2227 [==============================] - 0s 155us/sample - loss: 0.8714 - acc: 0.6331 - val_loss: 3.2000 - val_acc: 0.1221
Epoch 83/200
2227/2227 [==============================] - 0s 169us/sample - loss: 0.8791 - acc: 0.6255 - val_loss: 3.0979 - val_acc: 0.1221
Epoch 84/200
2227/2227 [==============================] - 0s 162us/sample - loss: 0.8871 - acc: 0.6273 - val_loss: 3.4153 - val_acc: 0.1221
Epoch 85/200
2227/2227 [==============================] - 0s 166us/sample - loss: 0.8923 - acc: 0.6228 - val_loss: 3.3456 - val_acc: 0.1185
Epoch 86/200
2227/2227 [==============================] - 0s 158us/sample - loss: 0.8758 - acc: 0.6354 - val_loss: 3.3381 - val_acc: 0.1167
Epoch 87/200
2227/2227 [==============================] - 0s 176us/sample - loss: 0.8711 - acc: 0.6448 - val_loss: 3.3348 - val_acc: 0.1239
Epoch 88/200
2227/2227 [==============================] - 0s 198us/sample - loss: 0.8810 - acc: 0.6421 - val_loss: 3.3381 - val_acc: 0.1005
Epoch 89/200
2227/2227 [==============================] - 0s 163us/sample - loss: 0.8891 - acc: 0.6331 - val_loss: 3.2295 - val_acc: 0.1185
Epoch 90/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8768 - acc: 0.6255 - val_loss: 3.0305 - val_acc: 0.1185
Epoch 91/200
2227/2227 [==============================] - 0s 159us/sample - loss: 0.8848 - acc: 0.6403 - val_loss: 3.1111 - val_acc: 0.1221
Epoch 92/200
2227/2227 [==============================] - 0s 158us/sample - loss: 0.8520 - acc: 0.6381 - val_loss: 3.1552 - val_acc: 0.1221
Epoch 93/200
2227/2227 [==============================] - 0s 164us/sample - loss: 0.8560 - acc: 0.6462 - val_loss: 3.1637 - val_acc: 0.1221
Epoch 94/200
2227/2227 [==============================] - 0s 166us/sample - loss: 0.8736 - acc: 0.6394 - val_loss: 3.0123 - val_acc: 0.1167
Epoch 95/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.8560 - acc: 0.6381 - val_loss: 2.9069 - val_acc: 0.1059
Epoch 96/200
2227/2227 [==============================] - 0s 166us/sample - loss: 0.8572 - acc: 0.6354 - val_loss: 2.9702 - val_acc: 0.1239
Epoch 97/200
2227/2227 [==============================] - 0s 169us/sample - loss: 0.8647 - acc: 0.6417 - val_loss: 3.0383 - val_acc: 0.1239
Epoch 98/200
2227/2227 [==============================] - 0s 176us/sample - loss: 0.8642 - acc: 0.6354 - val_loss: 3.2123 - val_acc: 0.1221
Epoch 99/200
2227/2227 [==============================] - 0s 167us/sample - loss: 0.8428 - acc: 0.6462 - val_loss: 3.1870 - val_acc: 0.1293
Epoch 100/200
2227/2227 [==============================] - 0s 175us/sample - loss: 0.8772 - acc: 0.6313 - val_loss: 3.1356 - val_acc: 0.1293
Epoch 101/200
2227/2227 [==============================] - 0s 170us/sample - loss: 0.8577 - acc: 0.6331 - val_loss: 3.1263 - val_acc: 0.1005
Epoch 102/200
2227/2227 [==============================] - 0s 194us/sample - loss: 0.8658 - acc: 0.6354 - val_loss: 3.1035 - val_acc: 0.1257
Epoch 103/200
2227/2227 [==============================] - 0s 181us/sample - loss: 0.8560 - acc: 0.6475 - val_loss: 3.1159 - val_acc: 0.1239
Epoch 104/200
2227/2227 [==============================] - 0s 170us/sample - loss: 0.8688 - acc: 0.6372 - val_loss: 3.0149 - val_acc: 0.1185
Epoch 105/200
2227/2227 [==============================] - 0s 190us/sample - loss: 0.8458 - acc: 0.6489 - val_loss: 3.0316 - val_acc: 0.1185
Epoch 106/200
2227/2227 [==============================] - 0s 192us/sample - loss: 0.8656 - acc: 0.6426 - val_loss: 3.0824 - val_acc: 0.1185
Epoch 107/200
2227/2227 [==============================] - 0s 220us/sample - loss: 0.8620 - acc: 0.6466 - val_loss: 2.9834 - val_acc: 0.1203
Epoch 108/200
2227/2227 [==============================] - 0s 189us/sample - loss: 0.8515 - acc: 0.6399 - val_loss: 2.9796 - val_acc: 0.1257
Epoch 109/200
2227/2227 [==============================] - 0s 156us/sample - loss: 0.8573 - acc: 0.6466 - val_loss: 3.1150 - val_acc: 0.1167
Epoch 110/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.8487 - acc: 0.6403 - val_loss: 3.0509 - val_acc: 0.1257
Epoch 111/200
2227/2227 [==============================] - 0s 150us/sample - loss: 0.8424 - acc: 0.6399 - val_loss: 3.0173 - val_acc: 0.1203
Epoch 112/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8398 - acc: 0.6399 - val_loss: 2.9130 - val_acc: 0.1077
Epoch 113/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.8391 - acc: 0.6556 - val_loss: 2.8357 - val_acc: 0.1167
Epoch 114/200
2227/2227 [==============================] - 0s 167us/sample - loss: 0.8306 - acc: 0.6471 - val_loss: 2.9391 - val_acc: 0.1203
Epoch 115/200
2227/2227 [==============================] - 0s 179us/sample - loss: 0.8407 - acc: 0.6408 - val_loss: 3.0049 - val_acc: 0.1257
Epoch 116/200
2227/2227 [==============================] - 0s 151us/sample - loss: 0.8454 - acc: 0.6340 - val_loss: 2.9501 - val_acc: 0.1239
Epoch 117/200
2227/2227 [==============================] - 0s 159us/sample - loss: 0.8428 - acc: 0.6484 - val_loss: 2.8989 - val_acc: 0.1239
Epoch 118/200
2227/2227 [==============================] - 0s 150us/sample - loss: 0.8299 - acc: 0.6502 - val_loss: 2.8866 - val_acc: 0.1257
Epoch 119/200
2227/2227 [==============================] - 0s 149us/sample - loss: 0.8310 - acc: 0.6596 - val_loss: 2.8615 - val_acc: 0.1239
Epoch 120/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8366 - acc: 0.6493 - val_loss: 2.9820 - val_acc: 0.1221
Epoch 121/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8372 - acc: 0.6489 - val_loss: 2.9054 - val_acc: 0.1185
Epoch 122/200
2227/2227 [==============================] - 0s 151us/sample - loss: 0.8539 - acc: 0.6363 - val_loss: 3.0316 - val_acc: 0.1185
Epoch 123/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8532 - acc: 0.6471 - val_loss: 2.8192 - val_acc: 0.1095
Epoch 124/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8533 - acc: 0.6426 - val_loss: 3.0235 - val_acc: 0.1149
Epoch 125/200
2227/2227 [==============================] - 0s 149us/sample - loss: 0.8287 - acc: 0.6682 - val_loss: 2.9674 - val_acc: 0.1293
Epoch 126/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8416 - acc: 0.6412 - val_loss: 2.8795 - val_acc: 0.1239
Epoch 127/200
2227/2227 [==============================] - 0s 149us/sample - loss: 0.8412 - acc: 0.6502 - val_loss: 2.9158 - val_acc: 0.1113
Epoch 128/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8264 - acc: 0.6547 - val_loss: 2.9904 - val_acc: 0.1293
Epoch 129/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8155 - acc: 0.6578 - val_loss: 3.0083 - val_acc: 0.1239
Epoch 130/200
2227/2227 [==============================] - 0s 177us/sample - loss: 0.8140 - acc: 0.6682 - val_loss: 3.0040 - val_acc: 0.1239
Epoch 131/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8225 - acc: 0.6646 - val_loss: 2.9898 - val_acc: 0.1275
Epoch 132/200
2227/2227 [==============================] - 0s 158us/sample - loss: 0.8162 - acc: 0.6641 - val_loss: 2.9370 - val_acc: 0.1257
Epoch 133/200
2227/2227 [==============================] - 0s 157us/sample - loss: 0.8160 - acc: 0.6637 - val_loss: 2.8954 - val_acc: 0.1005
Epoch 134/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8328 - acc: 0.6587 - val_loss: 3.0009 - val_acc: 0.1239
Epoch 135/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8273 - acc: 0.6587 - val_loss: 3.0034 - val_acc: 0.1311
Epoch 136/200
2227/2227 [==============================] - 0s 164us/sample - loss: 0.8290 - acc: 0.6551 - val_loss: 2.9249 - val_acc: 0.1275
Epoch 137/200
2227/2227 [==============================] - 0s 181us/sample - loss: 0.8316 - acc: 0.6493 - val_loss: 2.9559 - val_acc: 0.1041
Epoch 138/200
2227/2227 [==============================] - 0s 155us/sample - loss: 0.8217 - acc: 0.6605 - val_loss: 2.9895 - val_acc: 0.1293
Epoch 139/200
2227/2227 [==============================] - 0s 151us/sample - loss: 0.8249 - acc: 0.6480 - val_loss: 2.7943 - val_acc: 0.1275
Epoch 140/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.8146 - acc: 0.6610 - val_loss: 2.8363 - val_acc: 0.1239
Epoch 141/200
2227/2227 [==============================] - 0s 151us/sample - loss: 0.8217 - acc: 0.6587 - val_loss: 2.8709 - val_acc: 0.1167
Epoch 142/200
2227/2227 [==============================] - 0s 159us/sample - loss: 0.8185 - acc: 0.6569 - val_loss: 2.9683 - val_acc: 0.1293
Epoch 143/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8185 - acc: 0.6551 - val_loss: 2.8750 - val_acc: 0.1185
Epoch 144/200
2227/2227 [==============================] - 0s 165us/sample - loss: 0.8155 - acc: 0.6637 - val_loss: 2.8270 - val_acc: 0.1203
Epoch 145/200
2227/2227 [==============================] - 0s 172us/sample - loss: 0.8283 - acc: 0.6475 - val_loss: 2.8632 - val_acc: 0.1203
Epoch 146/200
2227/2227 [==============================] - 0s 150us/sample - loss: 0.8315 - acc: 0.6520 - val_loss: 2.6817 - val_acc: 0.1185
Epoch 147/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.8157 - acc: 0.6628 - val_loss: 2.8681 - val_acc: 0.1113
Epoch 148/200
2227/2227 [==============================] - 0s 157us/sample - loss: 0.8044 - acc: 0.6605 - val_loss: 2.9011 - val_acc: 0.1203
Epoch 149/200
2227/2227 [==============================] - 0s 151us/sample - loss: 0.8050 - acc: 0.6610 - val_loss: 2.9383 - val_acc: 0.1293
Epoch 150/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.8122 - acc: 0.6502 - val_loss: 2.8995 - val_acc: 0.1275
Epoch 151/200
2227/2227 [==============================] - 0s 157us/sample - loss: 0.8159 - acc: 0.6556 - val_loss: 2.7998 - val_acc: 0.1185
Epoch 152/200
2227/2227 [==============================] - 0s 172us/sample - loss: 0.7730 - acc: 0.6794 - val_loss: 2.8705 - val_acc: 0.1293
Epoch 153/200
2227/2227 [==============================] - 0s 176us/sample - loss: 0.8144 - acc: 0.6475 - val_loss: 2.8275 - val_acc: 0.1293
Epoch 154/200
2227/2227 [==============================] - 0s 161us/sample - loss: 0.7984 - acc: 0.6767 - val_loss: 2.8616 - val_acc: 0.1239
Epoch 155/200
2227/2227 [==============================] - 0s 155us/sample - loss: 0.8164 - acc: 0.6601 - val_loss: 2.9443 - val_acc: 0.1293
Epoch 156/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.8123 - acc: 0.6677 - val_loss: 2.6194 - val_acc: 0.1203
Epoch 157/200
2227/2227 [==============================] - 0s 156us/sample - loss: 0.8279 - acc: 0.6556 - val_loss: 2.7184 - val_acc: 0.1041
Epoch 158/200
2227/2227 [==============================] - 0s 148us/sample - loss: 0.8118 - acc: 0.6605 - val_loss: 2.8679 - val_acc: 0.1059
Epoch 159/200
2227/2227 [==============================] - 0s 186us/sample - loss: 0.8046 - acc: 0.6682 - val_loss: 2.8418 - val_acc: 0.1257
Epoch 160/200
2227/2227 [==============================] - 0s 173us/sample - loss: 0.8266 - acc: 0.6659 - val_loss: 2.7558 - val_acc: 0.1167
Epoch 161/200
2227/2227 [==============================] - 0s 177us/sample - loss: 0.8051 - acc: 0.6587 - val_loss: 2.8520 - val_acc: 0.1275
Epoch 162/200
2227/2227 [==============================] - 0s 176us/sample - loss: 0.7785 - acc: 0.6614 - val_loss: 2.8475 - val_acc: 0.1149
Epoch 163/200
2227/2227 [==============================] - 0s 176us/sample - loss: 0.8265 - acc: 0.6583 - val_loss: 2.7857 - val_acc: 0.1275
Epoch 164/200
2227/2227 [==============================] - 0s 185us/sample - loss: 0.8044 - acc: 0.6704 - val_loss: 2.8654 - val_acc: 0.1185
Epoch 165/200
2227/2227 [==============================] - 0s 193us/sample - loss: 0.8053 - acc: 0.6632 - val_loss: 2.8156 - val_acc: 0.1203
Epoch 166/200
2227/2227 [==============================] - 0s 173us/sample - loss: 0.7950 - acc: 0.6691 - val_loss: 2.7596 - val_acc: 0.1275
Epoch 167/200
2227/2227 [==============================] - 0s 163us/sample - loss: 0.7975 - acc: 0.6736 - val_loss: 2.8095 - val_acc: 0.1131
Epoch 168/200
2227/2227 [==============================] - 0s 151us/sample - loss: 0.7907 - acc: 0.6686 - val_loss: 2.8828 - val_acc: 0.1185
Epoch 169/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.8003 - acc: 0.6673 - val_loss: 2.8672 - val_acc: 0.1311
Epoch 170/200
2227/2227 [==============================] - 0s 164us/sample - loss: 0.7911 - acc: 0.6727 - val_loss: 2.8573 - val_acc: 0.1311
Epoch 171/200
2227/2227 [==============================] - 0s 165us/sample - loss: 0.7900 - acc: 0.6780 - val_loss: 2.9256 - val_acc: 0.1221
Epoch 172/200
2227/2227 [==============================] - 0s 207us/sample - loss: 0.7774 - acc: 0.6776 - val_loss: 2.9193 - val_acc: 0.1203
Epoch 173/200
2227/2227 [==============================] - 0s 180us/sample - loss: 0.8055 - acc: 0.6596 - val_loss: 2.8148 - val_acc: 0.0862
Epoch 174/200
2227/2227 [==============================] - 0s 179us/sample - loss: 0.8049 - acc: 0.6789 - val_loss: 2.9088 - val_acc: 0.1221
Epoch 175/200
2227/2227 [==============================] - 0s 198us/sample - loss: 0.8079 - acc: 0.6587 - val_loss: 2.9178 - val_acc: 0.1311
Epoch 176/200
2227/2227 [==============================] - 1s 291us/sample - loss: 0.7927 - acc: 0.6614 - val_loss: 2.8788 - val_acc: 0.1185
Epoch 177/200
2227/2227 [==============================] - 1s 400us/sample - loss: 0.8078 - acc: 0.6677 - val_loss: 2.9075 - val_acc: 0.1311
Epoch 178/200
2227/2227 [==============================] - 1s 342us/sample - loss: 0.7894 - acc: 0.6664 - val_loss: 2.8090 - val_acc: 0.1239
Epoch 179/200
2227/2227 [==============================] - 1s 407us/sample - loss: 0.8075 - acc: 0.6551 - val_loss: 2.8345 - val_acc: 0.1293
Epoch 180/200
2227/2227 [==============================] - 0s 179us/sample - loss: 0.7876 - acc: 0.6641 - val_loss: 2.9235 - val_acc: 0.1311
Epoch 181/200
2227/2227 [==============================] - 0s 222us/sample - loss: 0.8286 - acc: 0.6628 - val_loss: 2.9205 - val_acc: 0.1293
Epoch 182/200
2227/2227 [==============================] - 0s 181us/sample - loss: 0.7988 - acc: 0.6691 - val_loss: 2.8784 - val_acc: 0.1239
Epoch 183/200
2227/2227 [==============================] - 0s 189us/sample - loss: 0.8018 - acc: 0.6596 - val_loss: 2.8504 - val_acc: 0.1239
Epoch 184/200
2227/2227 [==============================] - 0s 185us/sample - loss: 0.7874 - acc: 0.6830 - val_loss: 2.8611 - val_acc: 0.1185
Epoch 185/200
2227/2227 [==============================] - 0s 181us/sample - loss: 0.7935 - acc: 0.6740 - val_loss: 2.8955 - val_acc: 0.1311
Epoch 186/200
2227/2227 [==============================] - 0s 175us/sample - loss: 0.7826 - acc: 0.6700 - val_loss: 2.8213 - val_acc: 0.1346
Epoch 187/200
2227/2227 [==============================] - 0s 178us/sample - loss: 0.8087 - acc: 0.6578 - val_loss: 2.8016 - val_acc: 0.1221
Epoch 188/200
2227/2227 [==============================] - 1s 234us/sample - loss: 0.7804 - acc: 0.6691 - val_loss: 2.9093 - val_acc: 0.1329
Epoch 189/200
2227/2227 [==============================] - 0s 190us/sample - loss: 0.7811 - acc: 0.6731 - val_loss: 2.9330 - val_acc: 0.1239
Epoch 190/200
2227/2227 [==============================] - 0s 172us/sample - loss: 0.8024 - acc: 0.6780 - val_loss: 2.7618 - val_acc: 0.1203
Epoch 191/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8023 - acc: 0.6704 - val_loss: 2.8070 - val_acc: 0.1221
Epoch 192/200
2227/2227 [==============================] - 0s 159us/sample - loss: 0.7928 - acc: 0.6619 - val_loss: 2.8738 - val_acc: 0.1257
Epoch 193/200
2227/2227 [==============================] - 0s 189us/sample - loss: 0.8010 - acc: 0.6807 - val_loss: 2.8967 - val_acc: 0.1185
Epoch 194/200
2227/2227 [==============================] - 0s 150us/sample - loss: 0.7853 - acc: 0.6767 - val_loss: 2.9613 - val_acc: 0.1346
Epoch 195/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.7692 - acc: 0.6713 - val_loss: 2.8624 - val_acc: 0.1329
Epoch 196/200
2227/2227 [==============================] - 0s 152us/sample - loss: 0.8088 - acc: 0.6686 - val_loss: 2.8228 - val_acc: 0.1221
Epoch 197/200
2227/2227 [==============================] - 0s 153us/sample - loss: 0.7747 - acc: 0.6727 - val_loss: 2.9361 - val_acc: 0.1329
Epoch 198/200
2227/2227 [==============================] - 0s 157us/sample - loss: 0.7799 - acc: 0.6740 - val_loss: 2.9793 - val_acc: 0.1329
Epoch 199/200
2227/2227 [==============================] - 0s 154us/sample - loss: 0.7815 - acc: 0.6731 - val_loss: 2.7779 - val_acc: 0.1329
Epoch 200/200
2227/2227 [==============================] - 0s 151us/sample - loss: 0.7658 - acc: 0.6731 - val_loss: 2.7771 - val_acc: 0.1221
    In [276]:
    print(history.history.keys())
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.show()

    dict_keys(['loss', 'acc', 'val_loss', 'val_acc'])
    In [289]:
    y_pred = model1.predict_classes(X_test)
y_classes = [np.argmax(y, axis=None, out=None) for y in y_test]
    In [290]:
    from sklearn.metrics import classification_report
print(classification_report(y_pred,y_classes))

                  precision    recall  f1-score   support

           3       0.00      0.00      0.00         5
           4       0.47      0.09      0.15       100
           5       0.47      0.66      0.55       156
           6       0.41      0.52      0.46       168
           7       0.61      0.43      0.51        99
           8       0.00      0.00      0.00         0

    accuracy                           0.46       528
   macro avg       0.33      0.28      0.28       528
weighted avg       0.48      0.46      0.43       528

    Training neural net on dataset obtained after applying Kbest algo

    In [389]:
    for column in Kbest_features.drop(columns=['Signal_Strength']).columns:
    Kbest_features[column] = Kbest_features[column].apply(lambda x:
                        (x - Kbest_features[column].mean())/Kbest_features[column].std()
                       )
    In [292]:
    categorical_labels = to_categorical(Kbest_features['Signal_Strength'], num_classes=None)
    In [293]:
    X = Kbest_features.drop(columns=['Signal_Strength']).values
Y = categorical_labels
X_train, X_test, y_train, y_test = train_test_split(X, Y,test_size=0.33, random_state=42)
    In [294]:
    model_k = Sequential()
model_k.add(Dense(32, input_dim=8, kernel_initializer='normal', activation='relu'))
model_k.add(Dropout(0.5))
model_k.add(Dense(64, activation='relu'))
model_k.add(Dropout(0.5))
model_k.add(Dense(9, activation='softmax'))
model_k.summary()

    Model: "sequential_24"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_81 (Dense)             (None, 32)                288       
_________________________________________________________________
dropout_57 (Dropout)         (None, 32)                0         
_________________________________________________________________
dense_82 (Dense)             (None, 64)                2112      
_________________________________________________________________
dropout_58 (Dropout)         (None, 64)                0         
_________________________________________________________________
dense_83 (Dense)             (None, 9)                 585       
=================================================================
Total params: 2,985
Trainable params: 2,985
Non-trainable params: 0
_________________________________________________________________
    In [295]:
    model_k.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    In [296]:
    history = model_k.fit(X_train, y_train, epochs=150, batch_size=10,  verbose=1, validation_split=0.2)

    Train on 856 samples, validate on 215 samples
Epoch 1/150
856/856 [==============================] - 1s 759us/sample - loss: 1.9172 - acc: 0.3832 - val_loss: 1.5205 - val_acc: 0.3767
Epoch 2/150
856/856 [==============================] - 0s 167us/sample - loss: 1.3300 - acc: 0.4626 - val_loss: 1.1418 - val_acc: 0.5953
Epoch 3/150
856/856 [==============================] - 0s 179us/sample - loss: 1.1827 - acc: 0.4977 - val_loss: 1.0345 - val_acc: 0.6372
Epoch 4/150
856/856 [==============================] - 0s 192us/sample - loss: 1.1280 - acc: 0.5491 - val_loss: 0.9867 - val_acc: 0.6326
Epoch 5/150
856/856 [==============================] - 0s 180us/sample - loss: 1.1016 - acc: 0.5549 - val_loss: 0.9572 - val_acc: 0.6558
Epoch 6/150
856/856 [==============================] - 0s 172us/sample - loss: 1.0473 - acc: 0.5783 - val_loss: 0.9303 - val_acc: 0.6512
Epoch 7/150
856/856 [==============================] - 0s 168us/sample - loss: 1.0909 - acc: 0.5502 - val_loss: 0.9269 - val_acc: 0.6558
Epoch 8/150
856/856 [==============================] - 0s 248us/sample - loss: 1.0707 - acc: 0.5409 - val_loss: 0.9102 - val_acc: 0.6698
Epoch 9/150
856/856 [==============================] - 0s 160us/sample - loss: 1.0194 - acc: 0.5666 - val_loss: 0.8943 - val_acc: 0.6698
Epoch 10/150
856/856 [==============================] - 0s 178us/sample - loss: 1.0707 - acc: 0.5549 - val_loss: 0.9067 - val_acc: 0.6512
Epoch 11/150
856/856 [==============================] - 0s 161us/sample - loss: 1.0461 - acc: 0.5584 - val_loss: 0.8996 - val_acc: 0.6698
Epoch 12/150
856/856 [==============================] - 0s 168us/sample - loss: 1.0311 - acc: 0.5736 - val_loss: 0.8940 - val_acc: 0.6791
Epoch 13/150
856/856 [==============================] - 0s 167us/sample - loss: 1.0348 - acc: 0.5619 - val_loss: 0.8997 - val_acc: 0.6698
Epoch 14/150
856/856 [==============================] - 0s 171us/sample - loss: 1.0303 - acc: 0.5771 - val_loss: 0.8944 - val_acc: 0.6837
Epoch 15/150
856/856 [==============================] - 0s 168us/sample - loss: 1.0154 - acc: 0.5864 - val_loss: 0.8950 - val_acc: 0.6744
Epoch 16/150
856/856 [==============================] - 0s 167us/sample - loss: 1.0052 - acc: 0.5876 - val_loss: 0.8869 - val_acc: 0.6837
Epoch 17/150
856/856 [==============================] - 0s 222us/sample - loss: 1.0233 - acc: 0.5864 - val_loss: 0.8843 - val_acc: 0.6884
Epoch 18/150
856/856 [==============================] - 0s 181us/sample - loss: 1.0195 - acc: 0.5654 - val_loss: 0.8816 - val_acc: 0.6791
Epoch 19/150
856/856 [==============================] - 0s 174us/sample - loss: 1.0156 - acc: 0.5876 - val_loss: 0.8796 - val_acc: 0.6605
Epoch 20/150
856/856 [==============================] - 0s 172us/sample - loss: 1.0091 - acc: 0.5829 - val_loss: 0.8827 - val_acc: 0.6698
Epoch 21/150
856/856 [==============================] - 0s 179us/sample - loss: 1.0218 - acc: 0.5806 - val_loss: 0.8825 - val_acc: 0.6744
Epoch 22/150
856/856 [==============================] - 0s 171us/sample - loss: 0.9988 - acc: 0.5935 - val_loss: 0.8855 - val_acc: 0.6651
Epoch 23/150
856/856 [==============================] - 0s 167us/sample - loss: 0.9996 - acc: 0.5993 - val_loss: 0.8699 - val_acc: 0.6651
Epoch 24/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9965 - acc: 0.5900 - val_loss: 0.8705 - val_acc: 0.6791
Epoch 25/150
856/856 [==============================] - 0s 214us/sample - loss: 1.0127 - acc: 0.5724 - val_loss: 0.8773 - val_acc: 0.6837
Epoch 26/150
856/856 [==============================] - 0s 197us/sample - loss: 1.0051 - acc: 0.5864 - val_loss: 0.8857 - val_acc: 0.6698
Epoch 27/150
856/856 [==============================] - 0s 168us/sample - loss: 0.9996 - acc: 0.5958 - val_loss: 0.8692 - val_acc: 0.6977
Epoch 28/150
856/856 [==============================] - 0s 172us/sample - loss: 1.0127 - acc: 0.5748 - val_loss: 0.8654 - val_acc: 0.6930
Epoch 29/150
856/856 [==============================] - 0s 199us/sample - loss: 0.9971 - acc: 0.5864 - val_loss: 0.8654 - val_acc: 0.6791
Epoch 30/150
856/856 [==============================] - 0s 173us/sample - loss: 1.0019 - acc: 0.5888 - val_loss: 0.8705 - val_acc: 0.6837
Epoch 31/150
856/856 [==============================] - 0s 181us/sample - loss: 1.0117 - acc: 0.5736 - val_loss: 0.8744 - val_acc: 0.6791
Epoch 32/150
856/856 [==============================] - 0s 197us/sample - loss: 0.9872 - acc: 0.5806 - val_loss: 0.8639 - val_acc: 0.6977
Epoch 33/150
856/856 [==============================] - 0s 186us/sample - loss: 1.0086 - acc: 0.6051 - val_loss: 0.8658 - val_acc: 0.6837
Epoch 34/150
856/856 [==============================] - 0s 178us/sample - loss: 0.9885 - acc: 0.5783 - val_loss: 0.8660 - val_acc: 0.6744
Epoch 35/150
856/856 [==============================] - 0s 175us/sample - loss: 0.9957 - acc: 0.6040 - val_loss: 0.8638 - val_acc: 0.6837
Epoch 36/150
856/856 [==============================] - 0s 159us/sample - loss: 0.9906 - acc: 0.5888 - val_loss: 0.8612 - val_acc: 0.6884
Epoch 37/150
856/856 [==============================] - 0s 169us/sample - loss: 1.0071 - acc: 0.6063 - val_loss: 0.8697 - val_acc: 0.6930
Epoch 38/150
856/856 [==============================] - 0s 164us/sample - loss: 0.9696 - acc: 0.5935 - val_loss: 0.8650 - val_acc: 0.6977
Epoch 39/150
856/856 [==============================] - 0s 179us/sample - loss: 0.9971 - acc: 0.5841 - val_loss: 0.8679 - val_acc: 0.6884
Epoch 40/150
856/856 [==============================] - 0s 167us/sample - loss: 0.9681 - acc: 0.6086 - val_loss: 0.8629 - val_acc: 0.7116
Epoch 41/150
856/856 [==============================] - 0s 271us/sample - loss: 0.9777 - acc: 0.5900 - val_loss: 0.8637 - val_acc: 0.6744
Epoch 42/150
856/856 [==============================] - 0s 202us/sample - loss: 0.9832 - acc: 0.5993 - val_loss: 0.8632 - val_acc: 0.6791
Epoch 43/150
856/856 [==============================] - 0s 169us/sample - loss: 0.9809 - acc: 0.6028 - val_loss: 0.8509 - val_acc: 0.6930
Epoch 44/150
856/856 [==============================] - 0s 174us/sample - loss: 0.9913 - acc: 0.6051 - val_loss: 0.8547 - val_acc: 0.6884
Epoch 45/150
856/856 [==============================] - 0s 173us/sample - loss: 0.9835 - acc: 0.5923 - val_loss: 0.8525 - val_acc: 0.6930
Epoch 46/150
856/856 [==============================] - 0s 186us/sample - loss: 0.9794 - acc: 0.6016 - val_loss: 0.8545 - val_acc: 0.6837
Epoch 47/150
856/856 [==============================] - 0s 304us/sample - loss: 0.9843 - acc: 0.5818 - val_loss: 0.8582 - val_acc: 0.6884
Epoch 48/150
856/856 [==============================] - 0s 190us/sample - loss: 0.9719 - acc: 0.5970 - val_loss: 0.8602 - val_acc: 0.6791
Epoch 49/150
856/856 [==============================] - 0s 204us/sample - loss: 0.9635 - acc: 0.5841 - val_loss: 0.8520 - val_acc: 0.6884
Epoch 50/150
856/856 [==============================] - 0s 196us/sample - loss: 0.9778 - acc: 0.5958 - val_loss: 0.8615 - val_acc: 0.6884
Epoch 51/150
856/856 [==============================] - 0s 236us/sample - loss: 0.9719 - acc: 0.5993 - val_loss: 0.8644 - val_acc: 0.6930
Epoch 52/150
856/856 [==============================] - 0s 179us/sample - loss: 0.9675 - acc: 0.5981 - val_loss: 0.8568 - val_acc: 0.6977
Epoch 53/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9749 - acc: 0.6028 - val_loss: 0.8608 - val_acc: 0.6930
Epoch 54/150
856/856 [==============================] - 0s 161us/sample - loss: 0.9569 - acc: 0.6075 - val_loss: 0.8532 - val_acc: 0.6930
Epoch 55/150
856/856 [==============================] - 0s 169us/sample - loss: 0.9605 - acc: 0.5923 - val_loss: 0.8535 - val_acc: 0.6930
Epoch 56/150
856/856 [==============================] - 0s 151us/sample - loss: 0.9529 - acc: 0.5853 - val_loss: 0.8463 - val_acc: 0.7023
Epoch 57/150
856/856 [==============================] - 0s 158us/sample - loss: 0.9651 - acc: 0.5841 - val_loss: 0.8498 - val_acc: 0.7116
Epoch 58/150
856/856 [==============================] - 0s 154us/sample - loss: 0.9638 - acc: 0.6075 - val_loss: 0.8567 - val_acc: 0.6977
Epoch 59/150
856/856 [==============================] - 0s 176us/sample - loss: 0.9710 - acc: 0.5888 - val_loss: 0.8528 - val_acc: 0.6744
Epoch 60/150
856/856 [==============================] - 0s 209us/sample - loss: 0.9706 - acc: 0.6051 - val_loss: 0.8541 - val_acc: 0.6977
Epoch 61/150
856/856 [==============================] - 0s 157us/sample - loss: 0.9567 - acc: 0.5993 - val_loss: 0.8493 - val_acc: 0.6837
Epoch 62/150
856/856 [==============================] - 0s 210us/sample - loss: 0.9625 - acc: 0.5958 - val_loss: 0.8413 - val_acc: 0.6977
Epoch 63/150
856/856 [==============================] - 0s 183us/sample - loss: 0.9482 - acc: 0.6016 - val_loss: 0.8442 - val_acc: 0.6930
Epoch 64/150
856/856 [==============================] - 0s 200us/sample - loss: 0.9584 - acc: 0.6005 - val_loss: 0.8424 - val_acc: 0.7023
Epoch 65/150
856/856 [==============================] - 0s 162us/sample - loss: 0.9620 - acc: 0.6005 - val_loss: 0.8546 - val_acc: 0.6791
Epoch 66/150
856/856 [==============================] - 0s 171us/sample - loss: 0.9620 - acc: 0.6051 - val_loss: 0.8455 - val_acc: 0.6930
Epoch 67/150
856/856 [==============================] - 0s 236us/sample - loss: 0.9387 - acc: 0.5946 - val_loss: 0.8442 - val_acc: 0.6930
Epoch 68/150
856/856 [==============================] - 0s 169us/sample - loss: 0.9639 - acc: 0.6063 - val_loss: 0.8511 - val_acc: 0.6884
Epoch 69/150
856/856 [==============================] - 0s 158us/sample - loss: 0.9636 - acc: 0.6028 - val_loss: 0.8508 - val_acc: 0.6977
Epoch 70/150
856/856 [==============================] - 0s 158us/sample - loss: 0.9642 - acc: 0.6005 - val_loss: 0.8567 - val_acc: 0.7023
Epoch 71/150
856/856 [==============================] - 0s 154us/sample - loss: 0.9707 - acc: 0.6063 - val_loss: 0.8618 - val_acc: 0.7070
Epoch 72/150
856/856 [==============================] - 0s 197us/sample - loss: 0.9647 - acc: 0.5911 - val_loss: 0.8699 - val_acc: 0.6558
Epoch 73/150
856/856 [==============================] - 0s 231us/sample - loss: 0.9780 - acc: 0.5876 - val_loss: 0.8608 - val_acc: 0.7023
Epoch 74/150
856/856 [==============================] - 0s 169us/sample - loss: 0.9581 - acc: 0.5946 - val_loss: 0.8608 - val_acc: 0.6884
Epoch 75/150
856/856 [==============================] - 0s 173us/sample - loss: 0.9587 - acc: 0.5864 - val_loss: 0.8623 - val_acc: 0.6837
Epoch 76/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9441 - acc: 0.5911 - val_loss: 0.8548 - val_acc: 0.6837
Epoch 77/150
856/856 [==============================] - 0s 180us/sample - loss: 0.9638 - acc: 0.5993 - val_loss: 0.8607 - val_acc: 0.6651
Epoch 78/150
856/856 [==============================] - 0s 169us/sample - loss: 0.9598 - acc: 0.6051 - val_loss: 0.8584 - val_acc: 0.6791
Epoch 79/150
856/856 [==============================] - 0s 193us/sample - loss: 0.9427 - acc: 0.6063 - val_loss: 0.8539 - val_acc: 0.6884
Epoch 80/150
856/856 [==============================] - 0s 174us/sample - loss: 0.9646 - acc: 0.5946 - val_loss: 0.8580 - val_acc: 0.6837
Epoch 81/150
856/856 [==============================] - 0s 168us/sample - loss: 0.9548 - acc: 0.6016 - val_loss: 0.8522 - val_acc: 0.6884
Epoch 82/150
856/856 [==============================] - 0s 181us/sample - loss: 0.9607 - acc: 0.5911 - val_loss: 0.8599 - val_acc: 0.6977
Epoch 83/150
856/856 [==============================] - 0s 183us/sample - loss: 0.9510 - acc: 0.6063 - val_loss: 0.8566 - val_acc: 0.6651
Epoch 84/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9485 - acc: 0.5958 - val_loss: 0.8548 - val_acc: 0.6744
Epoch 85/150
856/856 [==============================] - 0s 174us/sample - loss: 0.9541 - acc: 0.6075 - val_loss: 0.8524 - val_acc: 0.6977
Epoch 86/150
856/856 [==============================] - 0s 158us/sample - loss: 0.9591 - acc: 0.5970 - val_loss: 0.8544 - val_acc: 0.6930
Epoch 87/150
856/856 [==============================] - 0s 199us/sample - loss: 0.9381 - acc: 0.5981 - val_loss: 0.8563 - val_acc: 0.6884
Epoch 88/150
856/856 [==============================] - 0s 171us/sample - loss: 0.9443 - acc: 0.6121 - val_loss: 0.8565 - val_acc: 0.6744
Epoch 89/150
856/856 [==============================] - 0s 180us/sample - loss: 0.9700 - acc: 0.6005 - val_loss: 0.8537 - val_acc: 0.6791
Epoch 90/150
856/856 [==============================] - 0s 179us/sample - loss: 0.9353 - acc: 0.6121 - val_loss: 0.8536 - val_acc: 0.6791
Epoch 91/150
856/856 [==============================] - 0s 155us/sample - loss: 0.9404 - acc: 0.6063 - val_loss: 0.8495 - val_acc: 0.6791
Epoch 92/150
856/856 [==============================] - 0s 179us/sample - loss: 0.9416 - acc: 0.5981 - val_loss: 0.8492 - val_acc: 0.6930
Epoch 93/150
856/856 [==============================] - 0s 196us/sample - loss: 0.9558 - acc: 0.6086 - val_loss: 0.8501 - val_acc: 0.6977
Epoch 94/150
856/856 [==============================] - 0s 188us/sample - loss: 0.9410 - acc: 0.5935 - val_loss: 0.8515 - val_acc: 0.7023
Epoch 95/150
856/856 [==============================] - 0s 178us/sample - loss: 0.9452 - acc: 0.6040 - val_loss: 0.8516 - val_acc: 0.6884
Epoch 96/150
856/856 [==============================] - 0s 193us/sample - loss: 0.9532 - acc: 0.6028 - val_loss: 0.8553 - val_acc: 0.6884
Epoch 97/150
856/856 [==============================] - 0s 175us/sample - loss: 0.9234 - acc: 0.6145 - val_loss: 0.8441 - val_acc: 0.6791
Epoch 98/150
856/856 [==============================] - 0s 161us/sample - loss: 0.9327 - acc: 0.6180 - val_loss: 0.8452 - val_acc: 0.6930
Epoch 99/150
856/856 [==============================] - 0s 196us/sample - loss: 0.9548 - acc: 0.6343 - val_loss: 0.8442 - val_acc: 0.6930
Epoch 100/150
856/856 [==============================] - 0s 174us/sample - loss: 0.9453 - acc: 0.6133 - val_loss: 0.8474 - val_acc: 0.6977
Epoch 101/150
856/856 [==============================] - 0s 159us/sample - loss: 0.9457 - acc: 0.6145 - val_loss: 0.8520 - val_acc: 0.6837
Epoch 102/150
856/856 [==============================] - 0s 180us/sample - loss: 0.9471 - acc: 0.5900 - val_loss: 0.8552 - val_acc: 0.6744
Epoch 103/150
856/856 [==============================] - 0s 175us/sample - loss: 0.9476 - acc: 0.6121 - val_loss: 0.8531 - val_acc: 0.6977
Epoch 104/150
856/856 [==============================] - 0s 180us/sample - loss: 0.9348 - acc: 0.6016 - val_loss: 0.8406 - val_acc: 0.6930
Epoch 105/150
856/856 [==============================] - 0s 188us/sample - loss: 0.9496 - acc: 0.6016 - val_loss: 0.8431 - val_acc: 0.6930
Epoch 106/150
856/856 [==============================] - 0s 227us/sample - loss: 0.9560 - acc: 0.5993 - val_loss: 0.8475 - val_acc: 0.6837
Epoch 107/150
856/856 [==============================] - 0s 207us/sample - loss: 0.9467 - acc: 0.6063 - val_loss: 0.8521 - val_acc: 0.6698
Epoch 108/150
856/856 [==============================] - 0s 165us/sample - loss: 0.9459 - acc: 0.6098 - val_loss: 0.8498 - val_acc: 0.6837
Epoch 109/150
856/856 [==============================] - 0s 193us/sample - loss: 0.9380 - acc: 0.6110 - val_loss: 0.8481 - val_acc: 0.6837
Epoch 110/150
856/856 [==============================] - 0s 176us/sample - loss: 0.9190 - acc: 0.6157 - val_loss: 0.8486 - val_acc: 0.6837
Epoch 111/150
856/856 [==============================] - 0s 161us/sample - loss: 0.9436 - acc: 0.6180 - val_loss: 0.8523 - val_acc: 0.6837
Epoch 112/150
856/856 [==============================] - 0s 190us/sample - loss: 0.9437 - acc: 0.6121 - val_loss: 0.8510 - val_acc: 0.6744
Epoch 113/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9616 - acc: 0.5993 - val_loss: 0.8554 - val_acc: 0.6837
Epoch 114/150
856/856 [==============================] - 0s 169us/sample - loss: 0.9333 - acc: 0.6262 - val_loss: 0.8559 - val_acc: 0.6930
Epoch 115/150
856/856 [==============================] - 0s 189us/sample - loss: 0.9463 - acc: 0.6016 - val_loss: 0.8519 - val_acc: 0.6977
Epoch 116/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9216 - acc: 0.5993 - val_loss: 0.8495 - val_acc: 0.6930
Epoch 117/150
856/856 [==============================] - 0s 166us/sample - loss: 0.9199 - acc: 0.6227 - val_loss: 0.8449 - val_acc: 0.7023
Epoch 118/150
856/856 [==============================] - 0s 185us/sample - loss: 0.9562 - acc: 0.6075 - val_loss: 0.8557 - val_acc: 0.6791
Epoch 119/150
856/856 [==============================] - 0s 176us/sample - loss: 0.9398 - acc: 0.6110 - val_loss: 0.8461 - val_acc: 0.6744
Epoch 120/150
856/856 [==============================] - 0s 180us/sample - loss: 0.9427 - acc: 0.6180 - val_loss: 0.8496 - val_acc: 0.6791
Epoch 121/150
856/856 [==============================] - 0s 173us/sample - loss: 0.9376 - acc: 0.6086 - val_loss: 0.8506 - val_acc: 0.6744
Epoch 122/150
856/856 [==============================] - 0s 168us/sample - loss: 0.9443 - acc: 0.6180 - val_loss: 0.8426 - val_acc: 0.6837
Epoch 123/150
856/856 [==============================] - 0s 173us/sample - loss: 0.9215 - acc: 0.6262 - val_loss: 0.8454 - val_acc: 0.6837
Epoch 124/150
856/856 [==============================] - 0s 185us/sample - loss: 0.9168 - acc: 0.6086 - val_loss: 0.8461 - val_acc: 0.7023
Epoch 125/150
856/856 [==============================] - 0s 167us/sample - loss: 0.9398 - acc: 0.6110 - val_loss: 0.8507 - val_acc: 0.6930
Epoch 126/150
856/856 [==============================] - 0s 188us/sample - loss: 0.9453 - acc: 0.6063 - val_loss: 0.8475 - val_acc: 0.6837
Epoch 127/150
856/856 [==============================] - 0s 181us/sample - loss: 0.9431 - acc: 0.6051 - val_loss: 0.8504 - val_acc: 0.6744
Epoch 128/150
856/856 [==============================] - 0s 160us/sample - loss: 0.9341 - acc: 0.6086 - val_loss: 0.8450 - val_acc: 0.6884
Epoch 129/150
856/856 [==============================] - 0s 176us/sample - loss: 0.9528 - acc: 0.6051 - val_loss: 0.8559 - val_acc: 0.6698
Epoch 130/150
856/856 [==============================] - 0s 169us/sample - loss: 0.9355 - acc: 0.6005 - val_loss: 0.8533 - val_acc: 0.6744
Epoch 131/150
856/856 [==============================] - 0s 167us/sample - loss: 0.9119 - acc: 0.6285 - val_loss: 0.8485 - val_acc: 0.6884
Epoch 132/150
856/856 [==============================] - 0s 181us/sample - loss: 0.9585 - acc: 0.6133 - val_loss: 0.8486 - val_acc: 0.6791
Epoch 133/150
856/856 [==============================] - 0s 180us/sample - loss: 0.9499 - acc: 0.6086 - val_loss: 0.8436 - val_acc: 0.7023
Epoch 134/150
856/856 [==============================] - 0s 190us/sample - loss: 0.9415 - acc: 0.6110 - val_loss: 0.8492 - val_acc: 0.6884
Epoch 135/150
856/856 [==============================] - 0s 171us/sample - loss: 0.9117 - acc: 0.6168 - val_loss: 0.8455 - val_acc: 0.6930
Epoch 136/150
856/856 [==============================] - 0s 162us/sample - loss: 0.9155 - acc: 0.5981 - val_loss: 0.8507 - val_acc: 0.6698
Epoch 137/150
856/856 [==============================] - 0s 187us/sample - loss: 0.9273 - acc: 0.6098 - val_loss: 0.8464 - val_acc: 0.6837
Epoch 138/150
856/856 [==============================] - 0s 171us/sample - loss: 0.9230 - acc: 0.6133 - val_loss: 0.8513 - val_acc: 0.6791
Epoch 139/150
856/856 [==============================] - 0s 245us/sample - loss: 0.9465 - acc: 0.6086 - val_loss: 0.8528 - val_acc: 0.6837
Epoch 140/150
856/856 [==============================] - 0s 172us/sample - loss: 0.9302 - acc: 0.6192 - val_loss: 0.8593 - val_acc: 0.6791
Epoch 141/150
856/856 [==============================] - 0s 173us/sample - loss: 0.9278 - acc: 0.6308 - val_loss: 0.8493 - val_acc: 0.6744
Epoch 142/150
856/856 [==============================] - 0s 160us/sample - loss: 0.9254 - acc: 0.6285 - val_loss: 0.8480 - val_acc: 0.6698
Epoch 143/150
856/856 [==============================] - 0s 196us/sample - loss: 0.9280 - acc: 0.6157 - val_loss: 0.8510 - val_acc: 0.6744
Epoch 144/150
856/856 [==============================] - 0s 173us/sample - loss: 0.9275 - acc: 0.6145 - val_loss: 0.8548 - val_acc: 0.6698
Epoch 145/150
856/856 [==============================] - 0s 158us/sample - loss: 0.9259 - acc: 0.5970 - val_loss: 0.8573 - val_acc: 0.6837
Epoch 146/150
856/856 [==============================] - 0s 183us/sample - loss: 0.9397 - acc: 0.6168 - val_loss: 0.8618 - val_acc: 0.6698
Epoch 147/150
856/856 [==============================] - 0s 180us/sample - loss: 0.9288 - acc: 0.6215 - val_loss: 0.8615 - val_acc: 0.6837
Epoch 148/150
856/856 [==============================] - 0s 203us/sample - loss: 0.9164 - acc: 0.6098 - val_loss: 0.8476 - val_acc: 0.6744
Epoch 149/150
856/856 [==============================] - 0s 206us/sample - loss: 0.9429 - acc: 0.6086 - val_loss: 0.8579 - val_acc: 0.6791
Epoch 150/150
856/856 [==============================] - 0s 178us/sample - loss: 0.9221 - acc: 0.6086 - val_loss: 0.8534 - val_acc: 0.6744
    In [297]:
    print(history.history.keys())
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.show()

    dict_keys(['loss', 'acc', 'val_loss', 'val_acc'])
    In [299]:
    y_pred = model_k.predict_classes(X_test)
y_classes = [np.argmax(y, axis=None, out=None) for y in y_test]
    In [300]:
    from sklearn.metrics import classification_report
print(classification_report(y_pred,y_classes))

                  precision    recall  f1-score   support

           3       0.00      0.00      0.00         0
           4       0.00      0.00      0.00         0
           5       0.80      0.62      0.70       280
           6       0.47      0.54      0.50       188
           7       0.39      0.45      0.42        60
           8       0.00      0.00      0.00         0

    accuracy                           0.57       528
   macro avg       0.28      0.27      0.27       528
weighted avg       0.64      0.57      0.60       528

    Conclusion:

  • We tried to train our model on PCA dataset ,on features truncated using kbest algo and extratree classifier.
  • We also tried to balance our classes using overampling algorithm 'SMOTE'
  • All the above approaches failed to produce any significant improvement in the accuracy of our model.
  • Next we will try to apply some feature engineering and classify our dependent variable into 3 classes i.e high, medium and low.
  • The major of advantage of this approach would be that it would not only reduce the number of classes but will also accomodate the other signal strength values that are either not one of these numbers or are in decimal format.

    • Categorising the signal strengths

    In [356]:
    df2 = df1.copy()
    In [357]:
    def categorise_signal(row):
    if row >= 7:
        return 3
    elif row <=4:
        return 1
    else:
        return 2
df2['Signal_Strength'] = df2['Signal_Strength'].apply(categorise_signal)
    In [358]:
    df2['Signal_Strength'].value_counts()
    Out[358]:
    2    1319
3     217
1      63
Name: Signal_Strength, dtype: int64
    In [359]:
    sns.countplot(y='Signal_Strength',hue='Signal_Strength', data=df2)
    Out[359]:
    <matplotlib.axes._subplots.AxesSubplot at 0x1000b6688>
    In [367]:
    def create_model(input_dim,output_dim):
    model = Sequential()
    model.add(Dense(32, input_dim=input_dim, kernel_initializer='normal', activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(64, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(output_dim, activation='softmax'))
    print(model.summary())
    return model

    Training on Principal Components

    In [360]:
    finalDf['Signal_Strength'] = finalDf['Signal_Strength'].apply(categorise_signal)
sns.countplot(y='Signal_Strength',hue='Signal_Strength', data=finalDf)
    Out[360]:
    <matplotlib.axes._subplots.AxesSubplot at 0x1000b0bc8>
    In [361]:
    from keras.utils.np_utils import to_categorical
categorical_labels = to_categorical(finalDf['Signal_Strength'], num_classes=None)
xpca = finalDf.drop(columns=['Signal_Strength']).values
ypca = categorical_labels
X_train, X_test, y_train, y_test = train_test_split(xpca, ypca,test_size=0.33, random_state=42)
y_train_classes = [np.argmax(y, axis=None, out=None) for y in y_train]
sm = SMOTE(random_state=2)
X_train_res,y_train_res = sm.fit_sample(X_train,y_train_classes)
    In [380]:
    y_train_res = to_categorical(y_train_res,num_classes=None)

    training the model without upsampling

    In [370]:
    model = create_model(6,4)

    Model: "sequential_27"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_90 (Dense)             (None, 32)                224       
_________________________________________________________________
dropout_63 (Dropout)         (None, 32)                0         
_________________________________________________________________
dense_91 (Dense)             (None, 64)                2112      
_________________________________________________________________
dropout_64 (Dropout)         (None, 64)                0         
_________________________________________________________________
dense_92 (Dense)             (None, 4)                 260       
=================================================================
Total params: 2,596
Trainable params: 2,596
Non-trainable params: 0
_________________________________________________________________
None
    In [371]:
    model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
history = model.fit(X_train, y_train, epochs=150, batch_size=10,  verbose=1, validation_split=0.2)

    Train on 856 samples, validate on 215 samples
Epoch 1/150
856/856 [==============================] - 1s 815us/sample - loss: 1.1546 - acc: 0.6589 - val_loss: 0.7890 - val_acc: 0.8326
Epoch 2/150
856/856 [==============================] - 0s 202us/sample - loss: 0.6574 - acc: 0.8318 - val_loss: 0.5210 - val_acc: 0.8326
Epoch 3/150
856/856 [==============================] - 0s 209us/sample - loss: 0.5810 - acc: 0.8259 - val_loss: 0.4785 - val_acc: 0.8372
Epoch 4/150
856/856 [==============================] - 0s 164us/sample - loss: 0.5240 - acc: 0.8306 - val_loss: 0.4618 - val_acc: 0.8558
Epoch 5/150
856/856 [==============================] - 0s 238us/sample - loss: 0.4988 - acc: 0.8364 - val_loss: 0.4453 - val_acc: 0.8605
Epoch 6/150
856/856 [==============================] - 0s 162us/sample - loss: 0.5097 - acc: 0.8271 - val_loss: 0.4337 - val_acc: 0.8605
Epoch 7/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4758 - acc: 0.8470 - val_loss: 0.4222 - val_acc: 0.8558
Epoch 8/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4753 - acc: 0.8400 - val_loss: 0.4202 - val_acc: 0.8465
Epoch 9/150
856/856 [==============================] - 0s 171us/sample - loss: 0.4846 - acc: 0.8341 - val_loss: 0.4157 - val_acc: 0.8605
Epoch 10/150
856/856 [==============================] - 0s 227us/sample - loss: 0.4659 - acc: 0.8493 - val_loss: 0.4147 - val_acc: 0.8605
Epoch 11/150
856/856 [==============================] - 0s 167us/sample - loss: 0.4752 - acc: 0.8318 - val_loss: 0.4194 - val_acc: 0.8326
Epoch 12/150
856/856 [==============================] - 0s 167us/sample - loss: 0.4585 - acc: 0.8411 - val_loss: 0.4065 - val_acc: 0.8465
Epoch 13/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4713 - acc: 0.8329 - val_loss: 0.4164 - val_acc: 0.8558
Epoch 14/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4560 - acc: 0.8446 - val_loss: 0.4191 - val_acc: 0.8465
Epoch 15/150
856/856 [==============================] - 0s 221us/sample - loss: 0.4624 - acc: 0.8306 - val_loss: 0.4151 - val_acc: 0.8419
Epoch 16/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4637 - acc: 0.8400 - val_loss: 0.4064 - val_acc: 0.8512
Epoch 17/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4502 - acc: 0.8446 - val_loss: 0.4071 - val_acc: 0.8465
Epoch 18/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4422 - acc: 0.8423 - val_loss: 0.4062 - val_acc: 0.8605
Epoch 19/150
856/856 [==============================] - 0s 167us/sample - loss: 0.4402 - acc: 0.8388 - val_loss: 0.4044 - val_acc: 0.8558
Epoch 20/150
856/856 [==============================] - 0s 201us/sample - loss: 0.4262 - acc: 0.8446 - val_loss: 0.4105 - val_acc: 0.8419
Epoch 21/150
856/856 [==============================] - 0s 215us/sample - loss: 0.4504 - acc: 0.8329 - val_loss: 0.4045 - val_acc: 0.8512
Epoch 22/150
856/856 [==============================] - 0s 229us/sample - loss: 0.4527 - acc: 0.8435 - val_loss: 0.4011 - val_acc: 0.8465
Epoch 23/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4401 - acc: 0.8446 - val_loss: 0.4023 - val_acc: 0.8465
Epoch 24/150
856/856 [==============================] - 0s 165us/sample - loss: 0.4467 - acc: 0.8341 - val_loss: 0.4015 - val_acc: 0.8465
Epoch 25/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4452 - acc: 0.8376 - val_loss: 0.4019 - val_acc: 0.8651
Epoch 26/150
856/856 [==============================] - 0s 166us/sample - loss: 0.4541 - acc: 0.8423 - val_loss: 0.4050 - val_acc: 0.8558
Epoch 27/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4703 - acc: 0.8376 - val_loss: 0.4120 - val_acc: 0.8465
Epoch 28/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4505 - acc: 0.8353 - val_loss: 0.4113 - val_acc: 0.8512
Epoch 29/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4416 - acc: 0.8481 - val_loss: 0.4057 - val_acc: 0.8605
Epoch 30/150
856/856 [==============================] - 0s 167us/sample - loss: 0.4305 - acc: 0.8481 - val_loss: 0.4024 - val_acc: 0.8512
Epoch 31/150
856/856 [==============================] - 0s 176us/sample - loss: 0.4331 - acc: 0.8400 - val_loss: 0.4010 - val_acc: 0.8512
Epoch 32/150
856/856 [==============================] - 0s 194us/sample - loss: 0.4298 - acc: 0.8341 - val_loss: 0.3984 - val_acc: 0.8419
Epoch 33/150
856/856 [==============================] - 0s 224us/sample - loss: 0.4179 - acc: 0.8505 - val_loss: 0.3982 - val_acc: 0.8465
Epoch 34/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4355 - acc: 0.8376 - val_loss: 0.3994 - val_acc: 0.8512
Epoch 35/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4498 - acc: 0.8353 - val_loss: 0.4019 - val_acc: 0.8465
Epoch 36/150
856/856 [==============================] - 0s 174us/sample - loss: 0.4351 - acc: 0.8458 - val_loss: 0.3998 - val_acc: 0.8465
Epoch 37/150
856/856 [==============================] - 0s 203us/sample - loss: 0.4411 - acc: 0.8376 - val_loss: 0.4018 - val_acc: 0.8465
Epoch 38/150
856/856 [==============================] - 0s 190us/sample - loss: 0.4341 - acc: 0.8516 - val_loss: 0.3977 - val_acc: 0.8512
Epoch 39/150
856/856 [==============================] - 0s 196us/sample - loss: 0.4281 - acc: 0.8458 - val_loss: 0.3984 - val_acc: 0.8512
Epoch 40/150
856/856 [==============================] - 0s 207us/sample - loss: 0.4370 - acc: 0.8458 - val_loss: 0.4012 - val_acc: 0.8419
Epoch 41/150
856/856 [==============================] - 0s 164us/sample - loss: 0.4284 - acc: 0.8493 - val_loss: 0.4011 - val_acc: 0.8512
Epoch 42/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4343 - acc: 0.8446 - val_loss: 0.3980 - val_acc: 0.8605
Epoch 43/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4433 - acc: 0.8458 - val_loss: 0.3981 - val_acc: 0.8558
Epoch 44/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4244 - acc: 0.8458 - val_loss: 0.3981 - val_acc: 0.8512
Epoch 45/150
856/856 [==============================] - 0s 196us/sample - loss: 0.4222 - acc: 0.8470 - val_loss: 0.3989 - val_acc: 0.8372
Epoch 46/150
856/856 [==============================] - 0s 197us/sample - loss: 0.4267 - acc: 0.8481 - val_loss: 0.3988 - val_acc: 0.8512
Epoch 47/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4303 - acc: 0.8481 - val_loss: 0.3980 - val_acc: 0.8512
Epoch 48/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4332 - acc: 0.8400 - val_loss: 0.4002 - val_acc: 0.8512
Epoch 49/150
856/856 [==============================] - 0s 224us/sample - loss: 0.4117 - acc: 0.8493 - val_loss: 0.3998 - val_acc: 0.8465
Epoch 50/150
856/856 [==============================] - 0s 189us/sample - loss: 0.4093 - acc: 0.8528 - val_loss: 0.3960 - val_acc: 0.8512
Epoch 51/150
856/856 [==============================] - 0s 194us/sample - loss: 0.4126 - acc: 0.8528 - val_loss: 0.3990 - val_acc: 0.8419
Epoch 52/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4303 - acc: 0.8364 - val_loss: 0.3947 - val_acc: 0.8512
Epoch 53/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4278 - acc: 0.8458 - val_loss: 0.3984 - val_acc: 0.8512
Epoch 54/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4293 - acc: 0.8458 - val_loss: 0.3976 - val_acc: 0.8419
Epoch 55/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4240 - acc: 0.8470 - val_loss: 0.3983 - val_acc: 0.8465
Epoch 56/150
856/856 [==============================] - 0s 159us/sample - loss: 0.4194 - acc: 0.8458 - val_loss: 0.3996 - val_acc: 0.8465
Epoch 57/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4109 - acc: 0.8505 - val_loss: 0.3944 - val_acc: 0.8372
Epoch 58/150
856/856 [==============================] - 0s 174us/sample - loss: 0.4145 - acc: 0.8528 - val_loss: 0.3926 - val_acc: 0.8512
Epoch 59/150
856/856 [==============================] - 0s 185us/sample - loss: 0.4092 - acc: 0.8470 - val_loss: 0.3948 - val_acc: 0.8512
Epoch 60/150
856/856 [==============================] - 0s 164us/sample - loss: 0.4136 - acc: 0.8481 - val_loss: 0.3945 - val_acc: 0.8419
Epoch 61/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4208 - acc: 0.8435 - val_loss: 0.3967 - val_acc: 0.8465
Epoch 62/150
856/856 [==============================] - 0s 228us/sample - loss: 0.4415 - acc: 0.8423 - val_loss: 0.3978 - val_acc: 0.8465
Epoch 63/150
856/856 [==============================] - 0s 164us/sample - loss: 0.4108 - acc: 0.8435 - val_loss: 0.3915 - val_acc: 0.8512
Epoch 64/150
856/856 [==============================] - 0s 154us/sample - loss: 0.4105 - acc: 0.8540 - val_loss: 0.3945 - val_acc: 0.8465
Epoch 65/150
856/856 [==============================] - 0s 230us/sample - loss: 0.4190 - acc: 0.8376 - val_loss: 0.3906 - val_acc: 0.8419
Epoch 66/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4183 - acc: 0.8470 - val_loss: 0.3879 - val_acc: 0.8558
Epoch 67/150
856/856 [==============================] - 0s 157us/sample - loss: 0.4115 - acc: 0.8516 - val_loss: 0.3933 - val_acc: 0.8372
Epoch 68/150
856/856 [==============================] - 0s 167us/sample - loss: 0.4128 - acc: 0.8458 - val_loss: 0.3894 - val_acc: 0.8512
Epoch 69/150
856/856 [==============================] - 0s 154us/sample - loss: 0.4105 - acc: 0.8481 - val_loss: 0.3946 - val_acc: 0.8512
Epoch 70/150
856/856 [==============================] - 0s 290us/sample - loss: 0.4118 - acc: 0.8528 - val_loss: 0.3895 - val_acc: 0.8465
Epoch 71/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4257 - acc: 0.8423 - val_loss: 0.3946 - val_acc: 0.8512
Epoch 72/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4020 - acc: 0.8435 - val_loss: 0.3947 - val_acc: 0.8419
Epoch 73/150
856/856 [==============================] - 0s 165us/sample - loss: 0.4230 - acc: 0.8470 - val_loss: 0.3978 - val_acc: 0.8465
Epoch 74/150
856/856 [==============================] - 0s 162us/sample - loss: 0.4122 - acc: 0.8423 - val_loss: 0.3950 - val_acc: 0.8372
Epoch 75/150
856/856 [==============================] - 0s 160us/sample - loss: 0.4205 - acc: 0.8400 - val_loss: 0.3898 - val_acc: 0.8419
Epoch 76/150
856/856 [==============================] - 0s 158us/sample - loss: 0.4041 - acc: 0.8481 - val_loss: 0.3900 - val_acc: 0.8465
Epoch 77/150
856/856 [==============================] - 0s 159us/sample - loss: 0.4129 - acc: 0.8411 - val_loss: 0.3888 - val_acc: 0.8512
Epoch 78/150
856/856 [==============================] - 0s 165us/sample - loss: 0.4295 - acc: 0.8423 - val_loss: 0.3956 - val_acc: 0.8558
Epoch 79/150
856/856 [==============================] - 0s 162us/sample - loss: 0.4251 - acc: 0.8458 - val_loss: 0.3943 - val_acc: 0.8465
Epoch 80/150
856/856 [==============================] - 0s 161us/sample - loss: 0.4134 - acc: 0.8411 - val_loss: 0.3967 - val_acc: 0.8512
Epoch 81/150
856/856 [==============================] - 0s 164us/sample - loss: 0.4110 - acc: 0.8458 - val_loss: 0.3901 - val_acc: 0.8512
Epoch 82/150
856/856 [==============================] - 0s 175us/sample - loss: 0.4176 - acc: 0.8493 - val_loss: 0.3888 - val_acc: 0.8465
Epoch 83/150
856/856 [==============================] - 0s 210us/sample - loss: 0.4240 - acc: 0.8516 - val_loss: 0.3903 - val_acc: 0.8512
Epoch 84/150
856/856 [==============================] - 0s 209us/sample - loss: 0.4138 - acc: 0.8388 - val_loss: 0.3931 - val_acc: 0.8419
Epoch 85/150
856/856 [==============================] - 0s 200us/sample - loss: 0.4107 - acc: 0.8411 - val_loss: 0.3906 - val_acc: 0.8419
Epoch 86/150
856/856 [==============================] - 0s 197us/sample - loss: 0.3946 - acc: 0.8621 - val_loss: 0.3915 - val_acc: 0.8419
Epoch 87/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4012 - acc: 0.8470 - val_loss: 0.3921 - val_acc: 0.8419
Epoch 88/150
856/856 [==============================] - 0s 162us/sample - loss: 0.3880 - acc: 0.8505 - val_loss: 0.3878 - val_acc: 0.8512
Epoch 89/150
856/856 [==============================] - 0s 164us/sample - loss: 0.3967 - acc: 0.8586 - val_loss: 0.3907 - val_acc: 0.8465
Epoch 90/150
856/856 [==============================] - 0s 162us/sample - loss: 0.4089 - acc: 0.8528 - val_loss: 0.3902 - val_acc: 0.8512
Epoch 91/150
856/856 [==============================] - 0s 176us/sample - loss: 0.4141 - acc: 0.8400 - val_loss: 0.3900 - val_acc: 0.8558
Epoch 92/150
856/856 [==============================] - 0s 171us/sample - loss: 0.4066 - acc: 0.8540 - val_loss: 0.3904 - val_acc: 0.8419
Epoch 93/150
856/856 [==============================] - 0s 176us/sample - loss: 0.3972 - acc: 0.8458 - val_loss: 0.3928 - val_acc: 0.8558
Epoch 94/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4100 - acc: 0.8516 - val_loss: 0.3907 - val_acc: 0.8465
Epoch 95/150
856/856 [==============================] - 0s 174us/sample - loss: 0.4033 - acc: 0.8551 - val_loss: 0.3934 - val_acc: 0.8605
Epoch 96/150
856/856 [==============================] - 0s 176us/sample - loss: 0.4028 - acc: 0.8540 - val_loss: 0.3903 - val_acc: 0.8512
Epoch 97/150
856/856 [==============================] - 0s 174us/sample - loss: 0.4197 - acc: 0.8435 - val_loss: 0.3942 - val_acc: 0.8558
Epoch 98/150
856/856 [==============================] - 0s 174us/sample - loss: 0.4012 - acc: 0.8516 - val_loss: 0.3923 - val_acc: 0.8512
Epoch 99/150
856/856 [==============================] - 0s 165us/sample - loss: 0.4000 - acc: 0.8516 - val_loss: 0.3951 - val_acc: 0.8558
Epoch 100/150
856/856 [==============================] - 0s 171us/sample - loss: 0.3966 - acc: 0.8528 - val_loss: 0.3948 - val_acc: 0.8558
Epoch 101/150
856/856 [==============================] - 0s 242us/sample - loss: 0.3959 - acc: 0.8540 - val_loss: 0.3879 - val_acc: 0.8512
Epoch 102/150
856/856 [==============================] - 0s 265us/sample - loss: 0.4015 - acc: 0.8563 - val_loss: 0.3940 - val_acc: 0.8372
Epoch 103/150
856/856 [==============================] - 0s 252us/sample - loss: 0.4031 - acc: 0.8633 - val_loss: 0.3929 - val_acc: 0.8512
Epoch 104/150
856/856 [==============================] - 0s 204us/sample - loss: 0.3963 - acc: 0.8505 - val_loss: 0.3853 - val_acc: 0.8465
Epoch 105/150
856/856 [==============================] - 0s 166us/sample - loss: 0.3801 - acc: 0.8586 - val_loss: 0.3938 - val_acc: 0.8605
Epoch 106/150
856/856 [==============================] - 0s 180us/sample - loss: 0.3974 - acc: 0.8551 - val_loss: 0.3974 - val_acc: 0.8512
Epoch 107/150
856/856 [==============================] - 0s 161us/sample - loss: 0.4069 - acc: 0.8388 - val_loss: 0.3933 - val_acc: 0.8512
Epoch 108/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4038 - acc: 0.8505 - val_loss: 0.3926 - val_acc: 0.8465
Epoch 109/150
856/856 [==============================] - 0s 185us/sample - loss: 0.3899 - acc: 0.8551 - val_loss: 0.3955 - val_acc: 0.8512
Epoch 110/150
856/856 [==============================] - 0s 171us/sample - loss: 0.3883 - acc: 0.8575 - val_loss: 0.3904 - val_acc: 0.8558
Epoch 111/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4074 - acc: 0.8446 - val_loss: 0.3894 - val_acc: 0.8512
Epoch 112/150
856/856 [==============================] - 0s 185us/sample - loss: 0.3940 - acc: 0.8481 - val_loss: 0.3889 - val_acc: 0.8512
Epoch 113/150
856/856 [==============================] - 0s 178us/sample - loss: 0.4009 - acc: 0.8575 - val_loss: 0.3887 - val_acc: 0.8558
Epoch 114/150
856/856 [==============================] - 0s 175us/sample - loss: 0.3958 - acc: 0.8528 - val_loss: 0.3930 - val_acc: 0.8512
Epoch 115/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4002 - acc: 0.8528 - val_loss: 0.3894 - val_acc: 0.8558
Epoch 116/150
856/856 [==============================] - 0s 200us/sample - loss: 0.3981 - acc: 0.8493 - val_loss: 0.3908 - val_acc: 0.8419
Epoch 117/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4052 - acc: 0.8423 - val_loss: 0.3939 - val_acc: 0.8372
Epoch 118/150
856/856 [==============================] - 0s 185us/sample - loss: 0.3883 - acc: 0.8551 - val_loss: 0.3897 - val_acc: 0.8558
Epoch 119/150
856/856 [==============================] - 0s 188us/sample - loss: 0.3888 - acc: 0.8516 - val_loss: 0.3891 - val_acc: 0.8558
Epoch 120/150
856/856 [==============================] - 0s 169us/sample - loss: 0.3809 - acc: 0.8551 - val_loss: 0.3905 - val_acc: 0.8512
Epoch 121/150
856/856 [==============================] - 0s 165us/sample - loss: 0.3894 - acc: 0.8446 - val_loss: 0.3901 - val_acc: 0.8605
Epoch 122/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4069 - acc: 0.8563 - val_loss: 0.3871 - val_acc: 0.8651
Epoch 123/150
856/856 [==============================] - 0s 173us/sample - loss: 0.3918 - acc: 0.8563 - val_loss: 0.3863 - val_acc: 0.8744
Epoch 124/150
856/856 [==============================] - 0s 176us/sample - loss: 0.3945 - acc: 0.8516 - val_loss: 0.3886 - val_acc: 0.8651
Epoch 125/150
856/856 [==============================] - 0s 162us/sample - loss: 0.3871 - acc: 0.8598 - val_loss: 0.3886 - val_acc: 0.8512
Epoch 126/150
856/856 [==============================] - 0s 164us/sample - loss: 0.4055 - acc: 0.8435 - val_loss: 0.3909 - val_acc: 0.8465
Epoch 127/150
856/856 [==============================] - 0s 157us/sample - loss: 0.3960 - acc: 0.8505 - val_loss: 0.3875 - val_acc: 0.8651
Epoch 128/150
856/856 [==============================] - 0s 167us/sample - loss: 0.3786 - acc: 0.8645 - val_loss: 0.3938 - val_acc: 0.8558
Epoch 129/150
856/856 [==============================] - 0s 171us/sample - loss: 0.4063 - acc: 0.8446 - val_loss: 0.3908 - val_acc: 0.8558
Epoch 130/150
856/856 [==============================] - 0s 160us/sample - loss: 0.3875 - acc: 0.8540 - val_loss: 0.3917 - val_acc: 0.8512
Epoch 131/150
856/856 [==============================] - 0s 161us/sample - loss: 0.3831 - acc: 0.8481 - val_loss: 0.3898 - val_acc: 0.8512
Epoch 132/150
856/856 [==============================] - 0s 168us/sample - loss: 0.3903 - acc: 0.8505 - val_loss: 0.3874 - val_acc: 0.8512
Epoch 133/150
856/856 [==============================] - 0s 164us/sample - loss: 0.3751 - acc: 0.8505 - val_loss: 0.3844 - val_acc: 0.8512
Epoch 134/150
856/856 [==============================] - 0s 162us/sample - loss: 0.3793 - acc: 0.8692 - val_loss: 0.3930 - val_acc: 0.8512
Epoch 135/150
856/856 [==============================] - 0s 162us/sample - loss: 0.3929 - acc: 0.8598 - val_loss: 0.3892 - val_acc: 0.8558
Epoch 136/150
856/856 [==============================] - 0s 241us/sample - loss: 0.3783 - acc: 0.8610 - val_loss: 0.3879 - val_acc: 0.8465
Epoch 137/150
856/856 [==============================] - 0s 154us/sample - loss: 0.3861 - acc: 0.8516 - val_loss: 0.3881 - val_acc: 0.8558
Epoch 138/150
856/856 [==============================] - 0s 168us/sample - loss: 0.3784 - acc: 0.8610 - val_loss: 0.3926 - val_acc: 0.8419
Epoch 139/150
856/856 [==============================] - 0s 190us/sample - loss: 0.4029 - acc: 0.8528 - val_loss: 0.3911 - val_acc: 0.8512
Epoch 140/150
856/856 [==============================] - 0s 193us/sample - loss: 0.3915 - acc: 0.8423 - val_loss: 0.3927 - val_acc: 0.8465
Epoch 141/150
856/856 [==============================] - 0s 187us/sample - loss: 0.3819 - acc: 0.8586 - val_loss: 0.3983 - val_acc: 0.8512
Epoch 142/150
856/856 [==============================] - 0s 174us/sample - loss: 0.3888 - acc: 0.8575 - val_loss: 0.3925 - val_acc: 0.8558
Epoch 143/150
856/856 [==============================] - 0s 175us/sample - loss: 0.3718 - acc: 0.8610 - val_loss: 0.3914 - val_acc: 0.8512
Epoch 144/150
856/856 [==============================] - 0s 174us/sample - loss: 0.3851 - acc: 0.8645 - val_loss: 0.3939 - val_acc: 0.8605
Epoch 145/150
856/856 [==============================] - 0s 172us/sample - loss: 0.3770 - acc: 0.8551 - val_loss: 0.3926 - val_acc: 0.8512
Epoch 146/150
856/856 [==============================] - 0s 181us/sample - loss: 0.3919 - acc: 0.8586 - val_loss: 0.3967 - val_acc: 0.8465
Epoch 147/150
856/856 [==============================] - 0s 166us/sample - loss: 0.3901 - acc: 0.8551 - val_loss: 0.3922 - val_acc: 0.8465
Epoch 148/150
856/856 [==============================] - 0s 178us/sample - loss: 0.3821 - acc: 0.8586 - val_loss: 0.3892 - val_acc: 0.8512
Epoch 149/150
856/856 [==============================] - 0s 180us/sample - loss: 0.3872 - acc: 0.8505 - val_loss: 0.3943 - val_acc: 0.8419
Epoch 150/150
856/856 [==============================] - 0s 171us/sample - loss: 0.3915 - acc: 0.8575 - val_loss: 0.3948 - val_acc: 0.8465
    In [372]:
    print(history.history.keys())
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.show()

    dict_keys(['loss', 'acc', 'val_loss', 'val_acc'])
    In [374]:
    y_pred = model.predict_classes(X_test)
y_classes = [np.argmax(y, axis=None, out=None) for y in y_test]
print(classification_report(y_pred,y_classes))

                  precision    recall  f1-score   support

           1       0.00      0.00      0.00         0
           2       0.94      0.85      0.89       474
           3       0.36      0.52      0.43        54

    accuracy                           0.82       528
   macro avg       0.43      0.46      0.44       528
weighted avg       0.88      0.82      0.85       528

    Training after applying SMOTE

    In [382]:
    model_s = create_model(6,4)

    Model: "sequential_30"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_99 (Dense)             (None, 32)                224       
_________________________________________________________________
dropout_69 (Dropout)         (None, 32)                0         
_________________________________________________________________
dense_100 (Dense)            (None, 64)                2112      
_________________________________________________________________
dropout_70 (Dropout)         (None, 64)                0         
_________________________________________________________________
dense_101 (Dense)            (None, 4)                 260       
=================================================================
Total params: 2,596
Trainable params: 2,596
Non-trainable params: 0
_________________________________________________________________
None
    In [383]:
    model_s.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
history = model_s.fit(X_train_res, y_train_res, epochs=150, batch_size=10,  verbose=1, validation_split=0.2)

    Train on 2133 samples, validate on 534 samples
Epoch 1/150
2133/2133 [==============================] - 1s 410us/sample - loss: 1.0997 - acc: 0.4552 - val_loss: 1.0309 - val_acc: 0.6461
Epoch 2/150
2133/2133 [==============================] - 0s 161us/sample - loss: 0.8112 - acc: 0.6010 - val_loss: 0.7800 - val_acc: 0.7097
Epoch 3/150
2133/2133 [==============================] - 0s 171us/sample - loss: 0.7646 - acc: 0.6057 - val_loss: 0.7869 - val_acc: 0.6891
Epoch 4/150
2133/2133 [==============================] - 0s 158us/sample - loss: 0.7566 - acc: 0.6278 - val_loss: 0.7468 - val_acc: 0.7116
Epoch 5/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.7297 - acc: 0.6385 - val_loss: 0.7945 - val_acc: 0.7041
Epoch 6/150
2133/2133 [==============================] - 0s 159us/sample - loss: 0.7195 - acc: 0.6362 - val_loss: 0.7181 - val_acc: 0.7378
Epoch 7/150
2133/2133 [==============================] - 0s 180us/sample - loss: 0.6941 - acc: 0.6535 - val_loss: 0.7557 - val_acc: 0.7172
Epoch 8/150
2133/2133 [==============================] - 0s 197us/sample - loss: 0.6884 - acc: 0.6610 - val_loss: 0.6971 - val_acc: 0.7416
Epoch 9/150
2133/2133 [==============================] - 0s 165us/sample - loss: 0.6815 - acc: 0.6803 - val_loss: 0.7323 - val_acc: 0.7191
Epoch 10/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.6802 - acc: 0.6742 - val_loss: 0.6916 - val_acc: 0.7416
Epoch 11/150
2133/2133 [==============================] - 0s 164us/sample - loss: 0.6693 - acc: 0.6695 - val_loss: 0.6525 - val_acc: 0.7772
Epoch 12/150
2133/2133 [==============================] - 0s 232us/sample - loss: 0.6763 - acc: 0.6793 - val_loss: 0.6986 - val_acc: 0.7491
Epoch 13/150
2133/2133 [==============================] - 0s 193us/sample - loss: 0.6543 - acc: 0.6765 - val_loss: 0.6802 - val_acc: 0.7491
Epoch 14/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.6478 - acc: 0.6967 - val_loss: 0.7414 - val_acc: 0.7322
Epoch 15/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.6527 - acc: 0.6821 - val_loss: 0.7699 - val_acc: 0.7172
Epoch 16/150
2133/2133 [==============================] - 0s 175us/sample - loss: 0.6421 - acc: 0.7042 - val_loss: 0.7591 - val_acc: 0.7247
Epoch 17/150
2133/2133 [==============================] - 0s 177us/sample - loss: 0.6408 - acc: 0.6953 - val_loss: 0.7064 - val_acc: 0.7341
Epoch 18/150
2133/2133 [==============================] - 0s 223us/sample - loss: 0.6457 - acc: 0.6915 - val_loss: 0.6982 - val_acc: 0.7772
Epoch 19/150
2133/2133 [==============================] - 0s 216us/sample - loss: 0.6376 - acc: 0.7032 - val_loss: 0.7039 - val_acc: 0.7528
Epoch 20/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.6289 - acc: 0.6995 - val_loss: 0.5903 - val_acc: 0.8333
Epoch 21/150
2133/2133 [==============================] - 0s 171us/sample - loss: 0.6279 - acc: 0.7065 - val_loss: 0.6738 - val_acc: 0.7790
Epoch 22/150
2133/2133 [==============================] - 0s 163us/sample - loss: 0.6339 - acc: 0.6953 - val_loss: 0.7102 - val_acc: 0.7659
Epoch 23/150
2133/2133 [==============================] - 0s 168us/sample - loss: 0.6326 - acc: 0.7065 - val_loss: 0.6860 - val_acc: 0.7753
Epoch 24/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.6372 - acc: 0.6868 - val_loss: 0.6492 - val_acc: 0.7903
Epoch 25/150
2133/2133 [==============================] - 0s 196us/sample - loss: 0.6222 - acc: 0.7065 - val_loss: 0.7002 - val_acc: 0.7734
Epoch 26/150
2133/2133 [==============================] - 0s 170us/sample - loss: 0.6083 - acc: 0.7145 - val_loss: 0.6298 - val_acc: 0.7865
Epoch 27/150
2133/2133 [==============================] - 0s 169us/sample - loss: 0.6206 - acc: 0.7121 - val_loss: 0.6625 - val_acc: 0.7715
Epoch 28/150
2133/2133 [==============================] - 0s 195us/sample - loss: 0.6067 - acc: 0.7229 - val_loss: 0.7244 - val_acc: 0.7584
Epoch 29/150
2133/2133 [==============================] - 0s 165us/sample - loss: 0.6195 - acc: 0.7098 - val_loss: 0.7049 - val_acc: 0.7547
Epoch 30/150
2133/2133 [==============================] - 0s 181us/sample - loss: 0.6266 - acc: 0.7060 - val_loss: 0.7057 - val_acc: 0.7640
Epoch 31/150
2133/2133 [==============================] - 0s 181us/sample - loss: 0.5918 - acc: 0.7182 - val_loss: 0.7306 - val_acc: 0.7472
Epoch 32/150
2133/2133 [==============================] - 0s 174us/sample - loss: 0.6055 - acc: 0.7131 - val_loss: 0.6951 - val_acc: 0.7659
Epoch 33/150
2133/2133 [==============================] - 0s 163us/sample - loss: 0.6052 - acc: 0.7304 - val_loss: 0.6336 - val_acc: 0.7846
Epoch 34/150
2133/2133 [==============================] - 0s 203us/sample - loss: 0.6204 - acc: 0.7267 - val_loss: 0.6685 - val_acc: 0.7603
Epoch 35/150
2133/2133 [==============================] - 0s 199us/sample - loss: 0.6028 - acc: 0.7215 - val_loss: 0.6312 - val_acc: 0.7828
Epoch 36/150
2133/2133 [==============================] - 0s 174us/sample - loss: 0.6063 - acc: 0.7201 - val_loss: 0.7488 - val_acc: 0.7210
Epoch 37/150
2133/2133 [==============================] - 0s 171us/sample - loss: 0.6190 - acc: 0.7187 - val_loss: 0.6697 - val_acc: 0.7640
Epoch 38/150
2133/2133 [==============================] - 0s 198us/sample - loss: 0.5901 - acc: 0.7314 - val_loss: 0.7371 - val_acc: 0.7303
Epoch 39/150
2133/2133 [==============================] - 0s 175us/sample - loss: 0.6011 - acc: 0.7107 - val_loss: 0.6683 - val_acc: 0.7828
Epoch 40/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.5905 - acc: 0.7215 - val_loss: 0.6328 - val_acc: 0.7865
Epoch 41/150
2133/2133 [==============================] - 0s 200us/sample - loss: 0.5891 - acc: 0.7257 - val_loss: 0.6316 - val_acc: 0.7884
Epoch 42/150
2133/2133 [==============================] - 0s 161us/sample - loss: 0.5795 - acc: 0.7375 - val_loss: 0.6924 - val_acc: 0.7697
Epoch 43/150
2133/2133 [==============================] - 0s 160us/sample - loss: 0.6110 - acc: 0.7309 - val_loss: 0.6292 - val_acc: 0.7940
Epoch 44/150
2133/2133 [==============================] - 0s 172us/sample - loss: 0.5927 - acc: 0.7229 - val_loss: 0.7064 - val_acc: 0.7584
Epoch 45/150
2133/2133 [==============================] - 0s 165us/sample - loss: 0.5953 - acc: 0.7295 - val_loss: 0.6470 - val_acc: 0.7828
Epoch 46/150
2133/2133 [==============================] - 0s 170us/sample - loss: 0.5998 - acc: 0.7304 - val_loss: 0.6434 - val_acc: 0.7734
Epoch 47/150
2133/2133 [==============================] - 0s 193us/sample - loss: 0.6010 - acc: 0.7243 - val_loss: 0.6528 - val_acc: 0.7734
Epoch 48/150
2133/2133 [==============================] - 0s 176us/sample - loss: 0.5918 - acc: 0.7290 - val_loss: 0.6501 - val_acc: 0.7697
Epoch 49/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.5686 - acc: 0.7375 - val_loss: 0.5967 - val_acc: 0.7921
Epoch 50/150
2133/2133 [==============================] - 0s 174us/sample - loss: 0.5807 - acc: 0.7276 - val_loss: 0.6124 - val_acc: 0.7903
Epoch 51/150
2133/2133 [==============================] - 0s 174us/sample - loss: 0.6004 - acc: 0.7112 - val_loss: 0.6640 - val_acc: 0.7715
Epoch 52/150
2133/2133 [==============================] - 0s 184us/sample - loss: 0.5733 - acc: 0.7318 - val_loss: 0.6876 - val_acc: 0.7584
Epoch 53/150
2133/2133 [==============================] - 0s 181us/sample - loss: 0.5841 - acc: 0.7342 - val_loss: 0.6494 - val_acc: 0.7772
Epoch 54/150
2133/2133 [==============================] - 0s 163us/sample - loss: 0.5883 - acc: 0.7361 - val_loss: 0.6798 - val_acc: 0.7678
Epoch 55/150
2133/2133 [==============================] - 0s 172us/sample - loss: 0.5832 - acc: 0.7304 - val_loss: 0.6105 - val_acc: 0.8071
Epoch 56/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.5832 - acc: 0.7389 - val_loss: 0.6941 - val_acc: 0.7697
Epoch 57/150
2133/2133 [==============================] - 0s 167us/sample - loss: 0.5807 - acc: 0.7229 - val_loss: 0.6164 - val_acc: 0.8052
Epoch 58/150
2133/2133 [==============================] - 0s 195us/sample - loss: 0.5908 - acc: 0.7239 - val_loss: 0.6566 - val_acc: 0.7959
Epoch 59/150
2133/2133 [==============================] - 0s 185us/sample - loss: 0.5756 - acc: 0.7342 - val_loss: 0.7120 - val_acc: 0.7547
Epoch 60/150
2133/2133 [==============================] - 0s 200us/sample - loss: 0.5799 - acc: 0.7262 - val_loss: 0.6491 - val_acc: 0.7921
Epoch 61/150
2133/2133 [==============================] - 0s 180us/sample - loss: 0.5747 - acc: 0.7253 - val_loss: 0.6695 - val_acc: 0.7753
Epoch 62/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.5624 - acc: 0.7482 - val_loss: 0.6905 - val_acc: 0.7790
Epoch 63/150
2133/2133 [==============================] - 0s 171us/sample - loss: 0.5792 - acc: 0.7304 - val_loss: 0.6746 - val_acc: 0.7772
Epoch 64/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5763 - acc: 0.7304 - val_loss: 0.6191 - val_acc: 0.8090
Epoch 65/150
2133/2133 [==============================] - 0s 205us/sample - loss: 0.5717 - acc: 0.7459 - val_loss: 0.6879 - val_acc: 0.7584
Epoch 66/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5828 - acc: 0.7262 - val_loss: 0.6455 - val_acc: 0.7790
Epoch 67/150
2133/2133 [==============================] - 0s 163us/sample - loss: 0.5735 - acc: 0.7304 - val_loss: 0.5636 - val_acc: 0.8127
Epoch 68/150
2133/2133 [==============================] - 0s 170us/sample - loss: 0.5901 - acc: 0.7351 - val_loss: 0.6569 - val_acc: 0.7472
Epoch 69/150
2133/2133 [==============================] - 0s 161us/sample - loss: 0.5829 - acc: 0.7337 - val_loss: 0.6487 - val_acc: 0.7734
Epoch 70/150
2133/2133 [==============================] - 0s 167us/sample - loss: 0.5624 - acc: 0.7365 - val_loss: 0.6485 - val_acc: 0.7809
Epoch 71/150
2133/2133 [==============================] - 0s 170us/sample - loss: 0.5512 - acc: 0.7487 - val_loss: 0.6119 - val_acc: 0.8052
Epoch 72/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.5703 - acc: 0.7365 - val_loss: 0.6412 - val_acc: 0.7790
Epoch 73/150
2133/2133 [==============================] - 0s 178us/sample - loss: 0.5742 - acc: 0.7365 - val_loss: 0.6047 - val_acc: 0.8034
Epoch 74/150
2133/2133 [==============================] - 0s 167us/sample - loss: 0.5726 - acc: 0.7290 - val_loss: 0.6547 - val_acc: 0.7640
Epoch 75/150
2133/2133 [==============================] - 0s 164us/sample - loss: 0.5433 - acc: 0.7506 - val_loss: 0.6296 - val_acc: 0.7715
Epoch 76/150
2133/2133 [==============================] - 0s 167us/sample - loss: 0.5614 - acc: 0.7375 - val_loss: 0.6592 - val_acc: 0.7378
Epoch 77/150
2133/2133 [==============================] - 0s 181us/sample - loss: 0.5802 - acc: 0.7379 - val_loss: 0.6638 - val_acc: 0.7603
Epoch 78/150
2133/2133 [==============================] - 0s 179us/sample - loss: 0.5360 - acc: 0.7623 - val_loss: 0.6506 - val_acc: 0.7828
Epoch 79/150
2133/2133 [==============================] - 0s 199us/sample - loss: 0.5501 - acc: 0.7623 - val_loss: 0.6064 - val_acc: 0.7697
Epoch 80/150
2133/2133 [==============================] - 0s 161us/sample - loss: 0.5518 - acc: 0.7496 - val_loss: 0.6274 - val_acc: 0.7360
Epoch 81/150
2133/2133 [==============================] - 0s 170us/sample - loss: 0.5675 - acc: 0.7496 - val_loss: 0.6619 - val_acc: 0.7210
Epoch 82/150
2133/2133 [==============================] - 0s 174us/sample - loss: 0.5494 - acc: 0.7548 - val_loss: 0.6484 - val_acc: 0.7397
Epoch 83/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.5675 - acc: 0.7567 - val_loss: 0.6420 - val_acc: 0.7509
Epoch 84/150
2133/2133 [==============================] - 0s 182us/sample - loss: 0.5616 - acc: 0.7576 - val_loss: 0.6707 - val_acc: 0.7341
Epoch 85/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.5631 - acc: 0.7459 - val_loss: 0.6563 - val_acc: 0.7285
Epoch 86/150
2133/2133 [==============================] - 0s 165us/sample - loss: 0.5436 - acc: 0.7520 - val_loss: 0.6505 - val_acc: 0.7360
Epoch 87/150
2133/2133 [==============================] - 0s 168us/sample - loss: 0.5508 - acc: 0.7482 - val_loss: 0.5644 - val_acc: 0.8109
Epoch 88/150
2133/2133 [==============================] - 0s 174us/sample - loss: 0.5588 - acc: 0.7511 - val_loss: 0.6249 - val_acc: 0.7940
Epoch 89/150
2133/2133 [==============================] - 0s 174us/sample - loss: 0.5556 - acc: 0.7571 - val_loss: 0.6411 - val_acc: 0.7828
Epoch 90/150
2133/2133 [==============================] - 0s 175us/sample - loss: 0.5500 - acc: 0.7492 - val_loss: 0.7270 - val_acc: 0.6873
Epoch 91/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.5388 - acc: 0.7670 - val_loss: 0.6749 - val_acc: 0.7697
Epoch 92/150
2133/2133 [==============================] - 0s 176us/sample - loss: 0.5477 - acc: 0.7543 - val_loss: 0.6144 - val_acc: 0.8071
Epoch 93/150
2133/2133 [==============================] - 0s 222us/sample - loss: 0.5315 - acc: 0.7590 - val_loss: 0.6244 - val_acc: 0.8052
Epoch 94/150
2133/2133 [==============================] - 0s 161us/sample - loss: 0.5329 - acc: 0.7567 - val_loss: 0.6541 - val_acc: 0.7603
Epoch 95/150
2133/2133 [==============================] - 0s 158us/sample - loss: 0.5450 - acc: 0.7543 - val_loss: 0.6839 - val_acc: 0.7228
Epoch 96/150
2133/2133 [==============================] - 0s 180us/sample - loss: 0.5434 - acc: 0.7628 - val_loss: 0.6604 - val_acc: 0.7903
Epoch 97/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5316 - acc: 0.7501 - val_loss: 0.6613 - val_acc: 0.7884
Epoch 98/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5379 - acc: 0.7553 - val_loss: 0.5937 - val_acc: 0.8240
Epoch 99/150
2133/2133 [==============================] - 0s 184us/sample - loss: 0.5461 - acc: 0.7534 - val_loss: 0.6731 - val_acc: 0.7790
Epoch 100/150
2133/2133 [==============================] - 0s 161us/sample - loss: 0.5419 - acc: 0.7581 - val_loss: 0.5904 - val_acc: 0.8034
Epoch 101/150
2133/2133 [==============================] - 0s 191us/sample - loss: 0.5442 - acc: 0.7412 - val_loss: 0.6333 - val_acc: 0.7959
Epoch 102/150
2133/2133 [==============================] - 0s 181us/sample - loss: 0.5287 - acc: 0.7562 - val_loss: 0.6760 - val_acc: 0.7584
Epoch 103/150
2133/2133 [==============================] - 0s 223us/sample - loss: 0.5513 - acc: 0.7487 - val_loss: 0.6405 - val_acc: 0.7828
Epoch 104/150
2133/2133 [==============================] - 0s 187us/sample - loss: 0.5427 - acc: 0.7543 - val_loss: 0.6415 - val_acc: 0.8015
Epoch 105/150
2133/2133 [==============================] - 0s 170us/sample - loss: 0.5367 - acc: 0.7529 - val_loss: 0.5959 - val_acc: 0.7959
Epoch 106/150
2133/2133 [==============================] - 0s 198us/sample - loss: 0.5460 - acc: 0.7506 - val_loss: 0.6639 - val_acc: 0.7659
Epoch 107/150
2133/2133 [==============================] - 0s 177us/sample - loss: 0.5306 - acc: 0.7581 - val_loss: 0.6325 - val_acc: 0.7921
Epoch 108/150
2133/2133 [==============================] - 0s 178us/sample - loss: 0.5453 - acc: 0.7445 - val_loss: 0.6642 - val_acc: 0.7584
Epoch 109/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.5343 - acc: 0.7571 - val_loss: 0.6955 - val_acc: 0.7247
Epoch 110/150
2133/2133 [==============================] - 0s 203us/sample - loss: 0.5349 - acc: 0.7623 - val_loss: 0.7911 - val_acc: 0.5243
Epoch 111/150
2133/2133 [==============================] - 0s 218us/sample - loss: 0.5530 - acc: 0.7642 - val_loss: 0.6031 - val_acc: 0.8127
Epoch 112/150
2133/2133 [==============================] - 0s 188us/sample - loss: 0.5491 - acc: 0.7581 - val_loss: 0.6312 - val_acc: 0.7959
Epoch 113/150
2133/2133 [==============================] - 0s 186us/sample - loss: 0.5391 - acc: 0.7628 - val_loss: 0.7543 - val_acc: 0.6648
Epoch 114/150
2133/2133 [==============================] - 0s 181us/sample - loss: 0.5389 - acc: 0.7553 - val_loss: 0.5681 - val_acc: 0.8146
Epoch 115/150
2133/2133 [==============================] - 0s 177us/sample - loss: 0.5504 - acc: 0.7459 - val_loss: 0.6694 - val_acc: 0.7453
Epoch 116/150
2133/2133 [==============================] - 0s 194us/sample - loss: 0.5237 - acc: 0.7665 - val_loss: 0.6613 - val_acc: 0.7528
Epoch 117/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.5298 - acc: 0.7679 - val_loss: 0.6394 - val_acc: 0.7884
Epoch 118/150
2133/2133 [==============================] - 0s 183us/sample - loss: 0.5407 - acc: 0.7562 - val_loss: 0.6082 - val_acc: 0.8015
Epoch 119/150
2133/2133 [==============================] - 0s 179us/sample - loss: 0.5084 - acc: 0.7675 - val_loss: 0.5900 - val_acc: 0.8258
Epoch 120/150
2133/2133 [==============================] - 0s 159us/sample - loss: 0.5388 - acc: 0.7600 - val_loss: 0.6794 - val_acc: 0.7453
Epoch 121/150
2133/2133 [==============================] - 0s 159us/sample - loss: 0.5517 - acc: 0.7557 - val_loss: 0.6755 - val_acc: 0.7622
Epoch 122/150
2133/2133 [==============================] - 0s 157us/sample - loss: 0.5428 - acc: 0.7628 - val_loss: 0.6361 - val_acc: 0.8090
Epoch 123/150
2133/2133 [==============================] - 0s 165us/sample - loss: 0.5234 - acc: 0.7628 - val_loss: 0.6152 - val_acc: 0.8202
Epoch 124/150
2133/2133 [==============================] - 0s 161us/sample - loss: 0.5376 - acc: 0.7501 - val_loss: 0.6127 - val_acc: 0.8071
Epoch 125/150
2133/2133 [==============================] - 0s 153us/sample - loss: 0.5158 - acc: 0.7740 - val_loss: 0.6485 - val_acc: 0.7715
Epoch 126/150
2133/2133 [==============================] - 0s 158us/sample - loss: 0.5451 - acc: 0.7632 - val_loss: 0.5692 - val_acc: 0.8371
Epoch 127/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5408 - acc: 0.7506 - val_loss: 0.5829 - val_acc: 0.8390
Epoch 128/150
2133/2133 [==============================] - 0s 158us/sample - loss: 0.5379 - acc: 0.7567 - val_loss: 0.6709 - val_acc: 0.7772
Epoch 129/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5364 - acc: 0.7543 - val_loss: 0.6190 - val_acc: 0.7996
Epoch 130/150
2133/2133 [==============================] - 0s 160us/sample - loss: 0.5369 - acc: 0.7482 - val_loss: 0.6037 - val_acc: 0.8071
Epoch 131/150
2133/2133 [==============================] - 0s 157us/sample - loss: 0.5330 - acc: 0.7609 - val_loss: 0.5708 - val_acc: 0.8240
Epoch 132/150
2133/2133 [==============================] - 0s 156us/sample - loss: 0.5335 - acc: 0.7623 - val_loss: 0.7198 - val_acc: 0.7116
Epoch 133/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.5344 - acc: 0.7590 - val_loss: 0.6960 - val_acc: 0.7622
Epoch 134/150
2133/2133 [==============================] - 0s 166us/sample - loss: 0.5394 - acc: 0.7637 - val_loss: 0.6840 - val_acc: 0.7846
Epoch 135/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5394 - acc: 0.7670 - val_loss: 0.6978 - val_acc: 0.7640
Epoch 136/150
2133/2133 [==============================] - 0s 156us/sample - loss: 0.5327 - acc: 0.7595 - val_loss: 0.5825 - val_acc: 0.8315
Epoch 137/150
2133/2133 [==============================] - 0s 162us/sample - loss: 0.5274 - acc: 0.7623 - val_loss: 0.6093 - val_acc: 0.8240
Epoch 138/150
2133/2133 [==============================] - 0s 163us/sample - loss: 0.5020 - acc: 0.7736 - val_loss: 0.6225 - val_acc: 0.7921
Epoch 139/150
2133/2133 [==============================] - 0s 157us/sample - loss: 0.5136 - acc: 0.7684 - val_loss: 0.5893 - val_acc: 0.8146
Epoch 140/150
2133/2133 [==============================] - 0s 159us/sample - loss: 0.5434 - acc: 0.7684 - val_loss: 0.6160 - val_acc: 0.8202
Epoch 141/150
2133/2133 [==============================] - 0s 156us/sample - loss: 0.5330 - acc: 0.7679 - val_loss: 0.6335 - val_acc: 0.8127
Epoch 142/150
2133/2133 [==============================] - 0s 160us/sample - loss: 0.5392 - acc: 0.7586 - val_loss: 0.6418 - val_acc: 0.8052
Epoch 143/150
2133/2133 [==============================] - 0s 190us/sample - loss: 0.5173 - acc: 0.7750 - val_loss: 0.6038 - val_acc: 0.8090
Epoch 144/150
2133/2133 [==============================] - 0s 201us/sample - loss: 0.5292 - acc: 0.7754 - val_loss: 0.5869 - val_acc: 0.8127
Epoch 145/150
2133/2133 [==============================] - 0s 201us/sample - loss: 0.5301 - acc: 0.7595 - val_loss: 0.5945 - val_acc: 0.8240
Epoch 146/150
2133/2133 [==============================] - 1s 248us/sample - loss: 0.4982 - acc: 0.7839 - val_loss: 0.5702 - val_acc: 0.8333
Epoch 147/150
2133/2133 [==============================] - 0s 201us/sample - loss: 0.5062 - acc: 0.7782 - val_loss: 0.6148 - val_acc: 0.8202
Epoch 148/150
2133/2133 [==============================] - 0s 173us/sample - loss: 0.5265 - acc: 0.7689 - val_loss: 0.5926 - val_acc: 0.8296
Epoch 149/150
2133/2133 [==============================] - 0s 159us/sample - loss: 0.5393 - acc: 0.7637 - val_loss: 0.6849 - val_acc: 0.7734
Epoch 150/150
2133/2133 [==============================] - 0s 206us/sample - loss: 0.4950 - acc: 0.7806 - val_loss: 0.7064 - val_acc: 0.7434
    In [385]:
    print(history.history.keys())
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.show()

    dict_keys(['loss', 'acc', 'val_loss', 'val_acc'])
    In [386]:
    y_pred = model_s.predict_classes(X_test)
y_classes = [np.argmax(y, axis=None, out=None) for y in y_test]
print(classification_report(y_pred,y_classes))

                  precision    recall  f1-score   support

           1       0.52      0.09      0.16       119
           2       0.67      0.89      0.76       322
           3       0.55      0.48      0.51        87

    accuracy                           0.64       528
   macro avg       0.58      0.49      0.48       528
weighted avg       0.61      0.64      0.58       528

    After we applied feature engineering to categorise the target variables a bit more broadly the average weighted f1 score jumped up 0.85. Also the recall score in case of medium strength signal has impeoved alot. Also we do not need to balance the dataset as after apply SMOTE there was not sny improvement noticed

    In [390]:
    Kbest_features['Signal_Strength'] = Kbest_features['Signal_Strength'].apply(categorise_signal)
categorical_labels = to_categorical(Kbest_features['Signal_Strength'], num_classes=None)
X = Kbest_features.drop(columns=['Signal_Strength']).values
Y = categorical_labels
X_train, X_test, y_train, y_test = train_test_split(X, Y,test_size=0.33, random_state=42)
    In [395]:
    model_k = create_model(8,4)

    Model: "sequential_31"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_102 (Dense)            (None, 32)                288       
_________________________________________________________________
dropout_71 (Dropout)         (None, 32)                0         
_________________________________________________________________
dense_103 (Dense)            (None, 64)                2112      
_________________________________________________________________
dropout_72 (Dropout)         (None, 64)                0         
_________________________________________________________________
dense_104 (Dense)            (None, 4)                 260       
=================================================================
Total params: 2,660
Trainable params: 2,660
Non-trainable params: 0
_________________________________________________________________
None
    In [397]:
    model_k.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
history = model_k.fit(X_train, y_train, epochs=150, batch_size=10,  verbose=1, validation_split=0.2)

    Train on 856 samples, validate on 215 samples
Epoch 1/150
856/856 [==============================] - 1s 860us/sample - loss: 1.0180 - acc: 0.7547 - val_loss: 0.6404 - val_acc: 0.8326
Epoch 2/150
856/856 [==============================] - 0s 216us/sample - loss: 0.5949 - acc: 0.8294 - val_loss: 0.5275 - val_acc: 0.8326
Epoch 3/150
856/856 [==============================] - 0s 175us/sample - loss: 0.5337 - acc: 0.8353 - val_loss: 0.4949 - val_acc: 0.8326
Epoch 4/150
856/856 [==============================] - 0s 190us/sample - loss: 0.5141 - acc: 0.8329 - val_loss: 0.4734 - val_acc: 0.8372
Epoch 5/150
856/856 [==============================] - 0s 178us/sample - loss: 0.5042 - acc: 0.8259 - val_loss: 0.4574 - val_acc: 0.8419
Epoch 6/150
856/856 [==============================] - 0s 174us/sample - loss: 0.4839 - acc: 0.8166 - val_loss: 0.4414 - val_acc: 0.8558
Epoch 7/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4858 - acc: 0.8271 - val_loss: 0.4307 - val_acc: 0.8465
Epoch 8/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4724 - acc: 0.8423 - val_loss: 0.4235 - val_acc: 0.8512
Epoch 9/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4722 - acc: 0.8341 - val_loss: 0.4165 - val_acc: 0.8605
Epoch 10/150
856/856 [==============================] - 0s 234us/sample - loss: 0.4758 - acc: 0.8259 - val_loss: 0.4147 - val_acc: 0.8465
Epoch 11/150
856/856 [==============================] - 0s 190us/sample - loss: 0.4499 - acc: 0.8364 - val_loss: 0.4122 - val_acc: 0.8558
Epoch 12/150
856/856 [==============================] - 0s 192us/sample - loss: 0.4613 - acc: 0.8423 - val_loss: 0.4111 - val_acc: 0.8512
Epoch 13/150
856/856 [==============================] - 0s 245us/sample - loss: 0.4490 - acc: 0.8411 - val_loss: 0.4155 - val_acc: 0.8558
Epoch 14/150
856/856 [==============================] - 0s 190us/sample - loss: 0.4520 - acc: 0.8306 - val_loss: 0.4127 - val_acc: 0.8558
Epoch 15/150
856/856 [==============================] - 0s 185us/sample - loss: 0.4320 - acc: 0.8376 - val_loss: 0.4059 - val_acc: 0.8558
Epoch 16/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4705 - acc: 0.8423 - val_loss: 0.4025 - val_acc: 0.8558
Epoch 17/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4518 - acc: 0.8400 - val_loss: 0.4066 - val_acc: 0.8512
Epoch 18/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4564 - acc: 0.8329 - val_loss: 0.4126 - val_acc: 0.8465
Epoch 19/150
856/856 [==============================] - 0s 238us/sample - loss: 0.4459 - acc: 0.8446 - val_loss: 0.4146 - val_acc: 0.8419
Epoch 20/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4419 - acc: 0.8411 - val_loss: 0.4059 - val_acc: 0.8512
Epoch 21/150
856/856 [==============================] - 0s 192us/sample - loss: 0.4457 - acc: 0.8388 - val_loss: 0.4075 - val_acc: 0.8651
Epoch 22/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4360 - acc: 0.8411 - val_loss: 0.4056 - val_acc: 0.8558
Epoch 23/150
856/856 [==============================] - 0s 229us/sample - loss: 0.4409 - acc: 0.8470 - val_loss: 0.4041 - val_acc: 0.8605
Epoch 24/150
856/856 [==============================] - 0s 199us/sample - loss: 0.4566 - acc: 0.8400 - val_loss: 0.4072 - val_acc: 0.8605
Epoch 25/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4271 - acc: 0.8446 - val_loss: 0.4031 - val_acc: 0.8558
Epoch 26/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4267 - acc: 0.8400 - val_loss: 0.3984 - val_acc: 0.8651
Epoch 27/150
856/856 [==============================] - 0s 236us/sample - loss: 0.4362 - acc: 0.8423 - val_loss: 0.4015 - val_acc: 0.8651
Epoch 28/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4466 - acc: 0.8400 - val_loss: 0.4019 - val_acc: 0.8651
Epoch 29/150
856/856 [==============================] - 0s 176us/sample - loss: 0.4351 - acc: 0.8493 - val_loss: 0.4038 - val_acc: 0.8605
Epoch 30/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4191 - acc: 0.8505 - val_loss: 0.4025 - val_acc: 0.8605
Epoch 31/150
856/856 [==============================] - 0s 231us/sample - loss: 0.4294 - acc: 0.8446 - val_loss: 0.4015 - val_acc: 0.8605
Epoch 32/150
856/856 [==============================] - 0s 394us/sample - loss: 0.4452 - acc: 0.8341 - val_loss: 0.4019 - val_acc: 0.8558
Epoch 33/150
856/856 [==============================] - 0s 192us/sample - loss: 0.4320 - acc: 0.8423 - val_loss: 0.3981 - val_acc: 0.8698
Epoch 34/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4372 - acc: 0.8306 - val_loss: 0.3966 - val_acc: 0.8698
Epoch 35/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4343 - acc: 0.8470 - val_loss: 0.4017 - val_acc: 0.8605
Epoch 36/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4300 - acc: 0.8435 - val_loss: 0.3999 - val_acc: 0.8605
Epoch 37/150
856/856 [==============================] - 0s 190us/sample - loss: 0.4296 - acc: 0.8435 - val_loss: 0.4017 - val_acc: 0.8512
Epoch 38/150
856/856 [==============================] - 0s 228us/sample - loss: 0.4240 - acc: 0.8400 - val_loss: 0.3983 - val_acc: 0.8651
Epoch 39/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4217 - acc: 0.8435 - val_loss: 0.4018 - val_acc: 0.8605
Epoch 40/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4235 - acc: 0.8400 - val_loss: 0.4006 - val_acc: 0.8558
Epoch 41/150
856/856 [==============================] - 0s 192us/sample - loss: 0.4308 - acc: 0.8481 - val_loss: 0.4020 - val_acc: 0.8558
Epoch 42/150
856/856 [==============================] - 0s 235us/sample - loss: 0.4282 - acc: 0.8423 - val_loss: 0.4017 - val_acc: 0.8605
Epoch 43/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4201 - acc: 0.8435 - val_loss: 0.4017 - val_acc: 0.8512
Epoch 44/150
856/856 [==============================] - 0s 196us/sample - loss: 0.4161 - acc: 0.8470 - val_loss: 0.3950 - val_acc: 0.8698
Epoch 45/150
856/856 [==============================] - 0s 232us/sample - loss: 0.4115 - acc: 0.8481 - val_loss: 0.3992 - val_acc: 0.8605
Epoch 46/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4287 - acc: 0.8458 - val_loss: 0.3988 - val_acc: 0.8512
Epoch 47/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4304 - acc: 0.8411 - val_loss: 0.3935 - val_acc: 0.8558
Epoch 48/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4220 - acc: 0.8353 - val_loss: 0.3980 - val_acc: 0.8465
Epoch 49/150
856/856 [==============================] - 0s 188us/sample - loss: 0.4060 - acc: 0.8505 - val_loss: 0.3908 - val_acc: 0.8651
Epoch 50/150
856/856 [==============================] - 0s 206us/sample - loss: 0.4198 - acc: 0.8364 - val_loss: 0.3918 - val_acc: 0.8651
Epoch 51/150
856/856 [==============================] - 0s 235us/sample - loss: 0.4301 - acc: 0.8388 - val_loss: 0.3915 - val_acc: 0.8605
Epoch 52/150
856/856 [==============================] - 0s 189us/sample - loss: 0.4311 - acc: 0.8446 - val_loss: 0.3986 - val_acc: 0.8558
Epoch 53/150
856/856 [==============================] - 0s 196us/sample - loss: 0.4284 - acc: 0.8400 - val_loss: 0.3986 - val_acc: 0.8512
Epoch 54/150
856/856 [==============================] - 0s 189us/sample - loss: 0.4110 - acc: 0.8446 - val_loss: 0.3895 - val_acc: 0.8605
Epoch 55/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4155 - acc: 0.8481 - val_loss: 0.3909 - val_acc: 0.8651
Epoch 56/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4145 - acc: 0.8376 - val_loss: 0.3920 - val_acc: 0.8419
Epoch 57/150
856/856 [==============================] - 0s 178us/sample - loss: 0.4048 - acc: 0.8505 - val_loss: 0.3899 - val_acc: 0.8651
Epoch 58/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4154 - acc: 0.8411 - val_loss: 0.3960 - val_acc: 0.8558
Epoch 59/150
856/856 [==============================] - 0s 175us/sample - loss: 0.4295 - acc: 0.8388 - val_loss: 0.3940 - val_acc: 0.8512
Epoch 60/150
856/856 [==============================] - 0s 174us/sample - loss: 0.4226 - acc: 0.8516 - val_loss: 0.3979 - val_acc: 0.8512
Epoch 61/150
856/856 [==============================] - 0s 160us/sample - loss: 0.4153 - acc: 0.8575 - val_loss: 0.3882 - val_acc: 0.8605
Epoch 62/150
856/856 [==============================] - 0s 248us/sample - loss: 0.4337 - acc: 0.8388 - val_loss: 0.3874 - val_acc: 0.8605
Epoch 63/150
856/856 [==============================] - 0s 202us/sample - loss: 0.3934 - acc: 0.8516 - val_loss: 0.3882 - val_acc: 0.8605
Epoch 64/150
856/856 [==============================] - 0s 169us/sample - loss: 0.4061 - acc: 0.8481 - val_loss: 0.3862 - val_acc: 0.8558
Epoch 65/150
856/856 [==============================] - 0s 194us/sample - loss: 0.4099 - acc: 0.8435 - val_loss: 0.3902 - val_acc: 0.8605
Epoch 66/150
856/856 [==============================] - 0s 188us/sample - loss: 0.4169 - acc: 0.8446 - val_loss: 0.3924 - val_acc: 0.8558
Epoch 67/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4122 - acc: 0.8481 - val_loss: 0.3906 - val_acc: 0.8605
Epoch 68/150
856/856 [==============================] - 0s 166us/sample - loss: 0.4020 - acc: 0.8411 - val_loss: 0.3909 - val_acc: 0.8605
Epoch 69/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4132 - acc: 0.8458 - val_loss: 0.3878 - val_acc: 0.8558
Epoch 70/150
856/856 [==============================] - 0s 203us/sample - loss: 0.4039 - acc: 0.8446 - val_loss: 0.3889 - val_acc: 0.8605
Epoch 71/150
856/856 [==============================] - 0s 166us/sample - loss: 0.4044 - acc: 0.8575 - val_loss: 0.3893 - val_acc: 0.8558
Epoch 72/150
856/856 [==============================] - 0s 197us/sample - loss: 0.4159 - acc: 0.8364 - val_loss: 0.3897 - val_acc: 0.8512
Epoch 73/150
856/856 [==============================] - 0s 179us/sample - loss: 0.3879 - acc: 0.8493 - val_loss: 0.3844 - val_acc: 0.8605
Epoch 74/150
856/856 [==============================] - 0s 176us/sample - loss: 0.4210 - acc: 0.8458 - val_loss: 0.3886 - val_acc: 0.8605
Epoch 75/150
856/856 [==============================] - 0s 201us/sample - loss: 0.3986 - acc: 0.8458 - val_loss: 0.3821 - val_acc: 0.8605
Epoch 76/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4203 - acc: 0.8540 - val_loss: 0.3864 - val_acc: 0.8605
Epoch 77/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4123 - acc: 0.8423 - val_loss: 0.3881 - val_acc: 0.8651
Epoch 78/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4038 - acc: 0.8481 - val_loss: 0.3888 - val_acc: 0.8558
Epoch 79/150
856/856 [==============================] - 0s 196us/sample - loss: 0.4035 - acc: 0.8411 - val_loss: 0.3898 - val_acc: 0.8512
Epoch 80/150
856/856 [==============================] - 0s 186us/sample - loss: 0.3984 - acc: 0.8446 - val_loss: 0.3905 - val_acc: 0.8465
Epoch 81/150
856/856 [==============================] - 0s 206us/sample - loss: 0.4193 - acc: 0.8435 - val_loss: 0.3901 - val_acc: 0.8465
Epoch 82/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4017 - acc: 0.8446 - val_loss: 0.3940 - val_acc: 0.8512
Epoch 83/150
856/856 [==============================] - 0s 196us/sample - loss: 0.4054 - acc: 0.8505 - val_loss: 0.3918 - val_acc: 0.8512
Epoch 84/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4064 - acc: 0.8481 - val_loss: 0.3848 - val_acc: 0.8744
Epoch 85/150
856/856 [==============================] - 0s 187us/sample - loss: 0.3922 - acc: 0.8551 - val_loss: 0.3898 - val_acc: 0.8512
Epoch 86/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4033 - acc: 0.8435 - val_loss: 0.3869 - val_acc: 0.8465
Epoch 87/150
856/856 [==============================] - 0s 232us/sample - loss: 0.4043 - acc: 0.8563 - val_loss: 0.3870 - val_acc: 0.8558
Epoch 88/150
856/856 [==============================] - 0s 181us/sample - loss: 0.3989 - acc: 0.8400 - val_loss: 0.3951 - val_acc: 0.8512
Epoch 89/150
856/856 [==============================] - 0s 194us/sample - loss: 0.4006 - acc: 0.8376 - val_loss: 0.3817 - val_acc: 0.8651
Epoch 90/150
856/856 [==============================] - 0s 194us/sample - loss: 0.4157 - acc: 0.8411 - val_loss: 0.3863 - val_acc: 0.8605
Epoch 91/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4012 - acc: 0.8516 - val_loss: 0.3898 - val_acc: 0.8558
Epoch 92/150
856/856 [==============================] - 0s 199us/sample - loss: 0.4051 - acc: 0.8470 - val_loss: 0.3937 - val_acc: 0.8465
Epoch 93/150
856/856 [==============================] - 0s 235us/sample - loss: 0.3994 - acc: 0.8551 - val_loss: 0.3925 - val_acc: 0.8558
Epoch 94/150
856/856 [==============================] - 0s 207us/sample - loss: 0.4148 - acc: 0.8400 - val_loss: 0.3915 - val_acc: 0.8512
Epoch 95/150
856/856 [==============================] - 0s 188us/sample - loss: 0.3988 - acc: 0.8446 - val_loss: 0.3862 - val_acc: 0.8651
Epoch 96/150
856/856 [==============================] - 0s 194us/sample - loss: 0.4083 - acc: 0.8364 - val_loss: 0.3902 - val_acc: 0.8698
Epoch 97/150
856/856 [==============================] - 0s 178us/sample - loss: 0.3954 - acc: 0.8540 - val_loss: 0.3909 - val_acc: 0.8512
Epoch 98/150
856/856 [==============================] - 0s 217us/sample - loss: 0.4066 - acc: 0.8481 - val_loss: 0.3865 - val_acc: 0.8651
Epoch 99/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4062 - acc: 0.8516 - val_loss: 0.3856 - val_acc: 0.8651
Epoch 100/150
856/856 [==============================] - 0s 188us/sample - loss: 0.3964 - acc: 0.8540 - val_loss: 0.3924 - val_acc: 0.8558
Epoch 101/150
856/856 [==============================] - 0s 192us/sample - loss: 0.4097 - acc: 0.8353 - val_loss: 0.3880 - val_acc: 0.8605
Epoch 102/150
856/856 [==============================] - 0s 193us/sample - loss: 0.3941 - acc: 0.8540 - val_loss: 0.3903 - val_acc: 0.8605
Epoch 103/150
856/856 [==============================] - 0s 188us/sample - loss: 0.3851 - acc: 0.8563 - val_loss: 0.3894 - val_acc: 0.8512
Epoch 104/150
856/856 [==============================] - 0s 178us/sample - loss: 0.3981 - acc: 0.8516 - val_loss: 0.3907 - val_acc: 0.8605
Epoch 105/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4034 - acc: 0.8481 - val_loss: 0.3905 - val_acc: 0.8512
Epoch 106/150
856/856 [==============================] - 0s 265us/sample - loss: 0.3903 - acc: 0.8586 - val_loss: 0.3833 - val_acc: 0.8512
Epoch 107/150
856/856 [==============================] - 0s 190us/sample - loss: 0.4001 - acc: 0.8575 - val_loss: 0.3881 - val_acc: 0.8558
Epoch 108/150
856/856 [==============================] - 0s 180us/sample - loss: 0.3843 - acc: 0.8505 - val_loss: 0.3829 - val_acc: 0.8558
Epoch 109/150
856/856 [==============================] - 0s 179us/sample - loss: 0.3844 - acc: 0.8516 - val_loss: 0.3806 - val_acc: 0.8558
Epoch 110/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4027 - acc: 0.8411 - val_loss: 0.3859 - val_acc: 0.8605
Epoch 111/150
856/856 [==============================] - 0s 197us/sample - loss: 0.3972 - acc: 0.8598 - val_loss: 0.3866 - val_acc: 0.8512
Epoch 112/150
856/856 [==============================] - 0s 189us/sample - loss: 0.3922 - acc: 0.8551 - val_loss: 0.3820 - val_acc: 0.8605
Epoch 113/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4093 - acc: 0.8470 - val_loss: 0.3897 - val_acc: 0.8605
Epoch 114/150
856/856 [==============================] - 0s 186us/sample - loss: 0.3885 - acc: 0.8621 - val_loss: 0.3921 - val_acc: 0.8512
Epoch 115/150
856/856 [==============================] - 0s 176us/sample - loss: 0.3948 - acc: 0.8540 - val_loss: 0.3888 - val_acc: 0.8512
Epoch 116/150
856/856 [==============================] - 0s 183us/sample - loss: 0.3878 - acc: 0.8470 - val_loss: 0.3902 - val_acc: 0.8512
Epoch 117/150
856/856 [==============================] - 0s 203us/sample - loss: 0.3935 - acc: 0.8423 - val_loss: 0.3862 - val_acc: 0.8558
Epoch 118/150
856/856 [==============================] - 0s 174us/sample - loss: 0.3978 - acc: 0.8551 - val_loss: 0.3823 - val_acc: 0.8651
Epoch 119/150
856/856 [==============================] - 0s 232us/sample - loss: 0.3866 - acc: 0.8633 - val_loss: 0.3816 - val_acc: 0.8651
Epoch 120/150
856/856 [==============================] - 0s 227us/sample - loss: 0.3910 - acc: 0.8481 - val_loss: 0.3816 - val_acc: 0.8651
Epoch 121/150
856/856 [==============================] - 0s 175us/sample - loss: 0.3824 - acc: 0.8516 - val_loss: 0.3841 - val_acc: 0.8605
Epoch 122/150
856/856 [==============================] - 0s 178us/sample - loss: 0.3866 - acc: 0.8633 - val_loss: 0.3807 - val_acc: 0.8651
Epoch 123/150
856/856 [==============================] - 0s 171us/sample - loss: 0.3867 - acc: 0.8516 - val_loss: 0.3832 - val_acc: 0.8558
Epoch 124/150
856/856 [==============================] - 0s 236us/sample - loss: 0.3871 - acc: 0.8575 - val_loss: 0.3846 - val_acc: 0.8605
Epoch 125/150
856/856 [==============================] - 0s 204us/sample - loss: 0.4017 - acc: 0.8481 - val_loss: 0.3897 - val_acc: 0.8558
Epoch 126/150
856/856 [==============================] - 0s 208us/sample - loss: 0.3797 - acc: 0.8586 - val_loss: 0.3975 - val_acc: 0.8558
Epoch 127/150
856/856 [==============================] - 0s 203us/sample - loss: 0.3814 - acc: 0.8563 - val_loss: 0.3912 - val_acc: 0.8512
Epoch 128/150
856/856 [==============================] - 0s 201us/sample - loss: 0.3967 - acc: 0.8598 - val_loss: 0.3845 - val_acc: 0.8651
Epoch 129/150
856/856 [==============================] - 0s 174us/sample - loss: 0.3949 - acc: 0.8493 - val_loss: 0.3828 - val_acc: 0.8651
Epoch 130/150
856/856 [==============================] - 0s 187us/sample - loss: 0.3894 - acc: 0.8563 - val_loss: 0.3805 - val_acc: 0.8651
Epoch 131/150
856/856 [==============================] - 0s 243us/sample - loss: 0.3881 - acc: 0.8481 - val_loss: 0.3823 - val_acc: 0.8651
Epoch 132/150
856/856 [==============================] - 0s 175us/sample - loss: 0.3830 - acc: 0.8493 - val_loss: 0.3835 - val_acc: 0.8605
Epoch 133/150
856/856 [==============================] - 0s 165us/sample - loss: 0.3787 - acc: 0.8493 - val_loss: 0.3917 - val_acc: 0.8651
Epoch 134/150
856/856 [==============================] - 0s 161us/sample - loss: 0.3793 - acc: 0.8540 - val_loss: 0.3841 - val_acc: 0.8651
Epoch 135/150
856/856 [==============================] - 0s 166us/sample - loss: 0.3830 - acc: 0.8610 - val_loss: 0.3842 - val_acc: 0.8651
Epoch 136/150
856/856 [==============================] - 0s 169us/sample - loss: 0.3759 - acc: 0.8551 - val_loss: 0.3844 - val_acc: 0.8605
Epoch 137/150
856/856 [==============================] - 0s 172us/sample - loss: 0.3912 - acc: 0.8551 - val_loss: 0.3833 - val_acc: 0.8651
Epoch 138/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4045 - acc: 0.8540 - val_loss: 0.3823 - val_acc: 0.8651
Epoch 139/150
856/856 [==============================] - 0s 160us/sample - loss: 0.3874 - acc: 0.8516 - val_loss: 0.3884 - val_acc: 0.8558
Epoch 140/150
856/856 [==============================] - 0s 169us/sample - loss: 0.3821 - acc: 0.8610 - val_loss: 0.3869 - val_acc: 0.8605
Epoch 141/150
856/856 [==============================] - 0s 164us/sample - loss: 0.3793 - acc: 0.8540 - val_loss: 0.3882 - val_acc: 0.8558
Epoch 142/150
856/856 [==============================] - 0s 174us/sample - loss: 0.3862 - acc: 0.8563 - val_loss: 0.3864 - val_acc: 0.8651
Epoch 143/150
856/856 [==============================] - 0s 167us/sample - loss: 0.3782 - acc: 0.8633 - val_loss: 0.3821 - val_acc: 0.8651
Epoch 144/150
856/856 [==============================] - 0s 161us/sample - loss: 0.3749 - acc: 0.8540 - val_loss: 0.3842 - val_acc: 0.8651
Epoch 145/150
856/856 [==============================] - 0s 169us/sample - loss: 0.3993 - acc: 0.8505 - val_loss: 0.3834 - val_acc: 0.8605
Epoch 146/150
856/856 [==============================] - 0s 160us/sample - loss: 0.3713 - acc: 0.8563 - val_loss: 0.3850 - val_acc: 0.8651
Epoch 147/150
856/856 [==============================] - 0s 173us/sample - loss: 0.3826 - acc: 0.8540 - val_loss: 0.3854 - val_acc: 0.8651
Epoch 148/150
856/856 [==============================] - 0s 175us/sample - loss: 0.3735 - acc: 0.8586 - val_loss: 0.3897 - val_acc: 0.8558
Epoch 149/150
856/856 [==============================] - 0s 179us/sample - loss: 0.3783 - acc: 0.8610 - val_loss: 0.3848 - val_acc: 0.8605
Epoch 150/150
856/856 [==============================] - 0s 168us/sample - loss: 0.3631 - acc: 0.8668 - val_loss: 0.3855 - val_acc: 0.8605
    In [398]:
    y_pred = model_k.predict_classes(X_test)
y_classes = [np.argmax(y, axis=None, out=None) for y in y_test]
print(classification_report(y_pred,y_classes))

                  precision    recall  f1-score   support

           1       0.00      0.00      0.00         0
           2       0.94      0.84      0.89       482
           3       0.27      0.46      0.34        46

    accuracy                           0.81       528
   macro avg       0.40      0.43      0.41       528
weighted avg       0.88      0.81      0.84       528

    Training on normal dataset

    In [409]:
    df2 = df1.copy()
    In [412]:
    df2['Signal_Strength'] = df2['Signal_Strength'].apply(categorise_signal)
categorical_labels = to_categorical(df2['Signal_Strength'], num_classes=None)
X = df2.drop(columns=['Signal_Strength']).values
Y = categorical_labels
X_train, X_test, y_train, y_test = train_test_split(X, Y,test_size=0.33, random_state=42)
    In [414]:
    model1 = create_model(11,4)

    Model: "sequential_33"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_108 (Dense)            (None, 32)                384       
_________________________________________________________________
dropout_75 (Dropout)         (None, 32)                0         
_________________________________________________________________
dense_109 (Dense)            (None, 64)                2112      
_________________________________________________________________
dropout_76 (Dropout)         (None, 64)                0         
_________________________________________________________________
dense_110 (Dense)            (None, 4)                 260       
=================================================================
Total params: 2,756
Trainable params: 2,756
Non-trainable params: 0
_________________________________________________________________
None
    In [415]:
    model1.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
history = model1.fit(X_train, y_train, epochs=150, batch_size=10,  verbose=1, validation_split=0.2)

    Train on 856 samples, validate on 215 samples
Epoch 1/150
856/856 [==============================] - 1s 895us/sample - loss: 1.2093 - acc: 0.6671 - val_loss: 0.5764 - val_acc: 0.8326
Epoch 2/150
856/856 [==============================] - 0s 174us/sample - loss: 0.7859 - acc: 0.7967 - val_loss: 0.5984 - val_acc: 0.8326
Epoch 3/150
856/856 [==============================] - 0s 187us/sample - loss: 0.7189 - acc: 0.7991 - val_loss: 0.5484 - val_acc: 0.8326
Epoch 4/150
856/856 [==============================] - 0s 188us/sample - loss: 0.6817 - acc: 0.8178 - val_loss: 0.5352 - val_acc: 0.8326
Epoch 5/150
856/856 [==============================] - 0s 182us/sample - loss: 0.6157 - acc: 0.8283 - val_loss: 0.5457 - val_acc: 0.8326
Epoch 6/150
856/856 [==============================] - 0s 175us/sample - loss: 0.6046 - acc: 0.8283 - val_loss: 0.5451 - val_acc: 0.8326
Epoch 7/150
856/856 [==============================] - 0s 183us/sample - loss: 0.5852 - acc: 0.8294 - val_loss: 0.5252 - val_acc: 0.8326
Epoch 8/150
856/856 [==============================] - 0s 249us/sample - loss: 0.5917 - acc: 0.8294 - val_loss: 0.5239 - val_acc: 0.8326
Epoch 9/150
856/856 [==============================] - 0s 230us/sample - loss: 0.5855 - acc: 0.8283 - val_loss: 0.5239 - val_acc: 0.8326
Epoch 10/150
856/856 [==============================] - 0s 185us/sample - loss: 0.5807 - acc: 0.8294 - val_loss: 0.5288 - val_acc: 0.8326
Epoch 11/150
856/856 [==============================] - 0s 197us/sample - loss: 0.5702 - acc: 0.8294 - val_loss: 0.5298 - val_acc: 0.8326
Epoch 12/150
856/856 [==============================] - 0s 190us/sample - loss: 0.5811 - acc: 0.8283 - val_loss: 0.5302 - val_acc: 0.8326
Epoch 13/150
856/856 [==============================] - 0s 249us/sample - loss: 0.5646 - acc: 0.8294 - val_loss: 0.5329 - val_acc: 0.8326
Epoch 14/150
856/856 [==============================] - 0s 186us/sample - loss: 0.5678 - acc: 0.8283 - val_loss: 0.5164 - val_acc: 0.8326
Epoch 15/150
856/856 [==============================] - 0s 197us/sample - loss: 0.5579 - acc: 0.8294 - val_loss: 0.5190 - val_acc: 0.8326
Epoch 16/150
856/856 [==============================] - 0s 232us/sample - loss: 0.5438 - acc: 0.8294 - val_loss: 0.5289 - val_acc: 0.8326
Epoch 17/150
856/856 [==============================] - 0s 243us/sample - loss: 0.5576 - acc: 0.8283 - val_loss: 0.5346 - val_acc: 0.8326
Epoch 18/150
856/856 [==============================] - 0s 188us/sample - loss: 0.5594 - acc: 0.8283 - val_loss: 0.5290 - val_acc: 0.8326
Epoch 19/150
856/856 [==============================] - 0s 194us/sample - loss: 0.5194 - acc: 0.8318 - val_loss: 0.5120 - val_acc: 0.8326
Epoch 20/150
856/856 [==============================] - 0s 183us/sample - loss: 0.5522 - acc: 0.8306 - val_loss: 0.5204 - val_acc: 0.8326
Epoch 21/150
856/856 [==============================] - 0s 187us/sample - loss: 0.5446 - acc: 0.8271 - val_loss: 0.5289 - val_acc: 0.8326
Epoch 22/150
856/856 [==============================] - 0s 179us/sample - loss: 0.5294 - acc: 0.8306 - val_loss: 0.5108 - val_acc: 0.8326
Epoch 23/150
856/856 [==============================] - 0s 179us/sample - loss: 0.5366 - acc: 0.8283 - val_loss: 0.5142 - val_acc: 0.8326
Epoch 24/150
856/856 [==============================] - 0s 179us/sample - loss: 0.5445 - acc: 0.8271 - val_loss: 0.5159 - val_acc: 0.8326
Epoch 25/150
856/856 [==============================] - 0s 211us/sample - loss: 0.5280 - acc: 0.8294 - val_loss: 0.5108 - val_acc: 0.8326
Epoch 26/150
856/856 [==============================] - 0s 284us/sample - loss: 0.5438 - acc: 0.8294 - val_loss: 0.5110 - val_acc: 0.8326
Epoch 27/150
856/856 [==============================] - 0s 182us/sample - loss: 0.5388 - acc: 0.8271 - val_loss: 0.5092 - val_acc: 0.8326
Epoch 28/150
856/856 [==============================] - 0s 187us/sample - loss: 0.5280 - acc: 0.8294 - val_loss: 0.5174 - val_acc: 0.8326
Epoch 29/150
856/856 [==============================] - 0s 194us/sample - loss: 0.5330 - acc: 0.8294 - val_loss: 0.5118 - val_acc: 0.8326
Epoch 30/150
856/856 [==============================] - 0s 187us/sample - loss: 0.5309 - acc: 0.8294 - val_loss: 0.5108 - val_acc: 0.8326
Epoch 31/150
856/856 [==============================] - 0s 195us/sample - loss: 0.5225 - acc: 0.8294 - val_loss: 0.5081 - val_acc: 0.8326
Epoch 32/150
856/856 [==============================] - 0s 202us/sample - loss: 0.5323 - acc: 0.8271 - val_loss: 0.5131 - val_acc: 0.8326
Epoch 33/150
856/856 [==============================] - 0s 182us/sample - loss: 0.5257 - acc: 0.8306 - val_loss: 0.4980 - val_acc: 0.8326
Epoch 34/150
856/856 [==============================] - 0s 183us/sample - loss: 0.5274 - acc: 0.8329 - val_loss: 0.5105 - val_acc: 0.8326
Epoch 35/150
856/856 [==============================] - 0s 188us/sample - loss: 0.5389 - acc: 0.8294 - val_loss: 0.5052 - val_acc: 0.8326
Epoch 36/150
856/856 [==============================] - 0s 218us/sample - loss: 0.5258 - acc: 0.8294 - val_loss: 0.5061 - val_acc: 0.8326
Epoch 37/150
856/856 [==============================] - 0s 239us/sample - loss: 0.5325 - acc: 0.8306 - val_loss: 0.5034 - val_acc: 0.8326
Epoch 38/150
856/856 [==============================] - 0s 243us/sample - loss: 0.5302 - acc: 0.8306 - val_loss: 0.5044 - val_acc: 0.8326
Epoch 39/150
856/856 [==============================] - 0s 201us/sample - loss: 0.5250 - acc: 0.8341 - val_loss: 0.4993 - val_acc: 0.8326
Epoch 40/150
856/856 [==============================] - 0s 247us/sample - loss: 0.5313 - acc: 0.8283 - val_loss: 0.4969 - val_acc: 0.8326
Epoch 41/150
856/856 [==============================] - 0s 217us/sample - loss: 0.5288 - acc: 0.8329 - val_loss: 0.4853 - val_acc: 0.8326
Epoch 42/150
856/856 [==============================] - 0s 307us/sample - loss: 0.5308 - acc: 0.8306 - val_loss: 0.5002 - val_acc: 0.8326
Epoch 43/150
856/856 [==============================] - 0s 193us/sample - loss: 0.5245 - acc: 0.8318 - val_loss: 0.5103 - val_acc: 0.8326
Epoch 44/150
856/856 [==============================] - 0s 179us/sample - loss: 0.5211 - acc: 0.8318 - val_loss: 0.4838 - val_acc: 0.8326
Epoch 45/150
856/856 [==============================] - 0s 195us/sample - loss: 0.5259 - acc: 0.8294 - val_loss: 0.4892 - val_acc: 0.8326
Epoch 46/150
856/856 [==============================] - 0s 181us/sample - loss: 0.5275 - acc: 0.8329 - val_loss: 0.4967 - val_acc: 0.8326
Epoch 47/150
856/856 [==============================] - 0s 183us/sample - loss: 0.5238 - acc: 0.8294 - val_loss: 0.4948 - val_acc: 0.8326
Epoch 48/150
856/856 [==============================] - 0s 186us/sample - loss: 0.5105 - acc: 0.8306 - val_loss: 0.4835 - val_acc: 0.8326
Epoch 49/150
856/856 [==============================] - 0s 239us/sample - loss: 0.5218 - acc: 0.8271 - val_loss: 0.4899 - val_acc: 0.8326
Epoch 50/150
856/856 [==============================] - 0s 188us/sample - loss: 0.5057 - acc: 0.8306 - val_loss: 0.4801 - val_acc: 0.8326
Epoch 51/150
856/856 [==============================] - 0s 174us/sample - loss: 0.5138 - acc: 0.8329 - val_loss: 0.4936 - val_acc: 0.8326
Epoch 52/150
856/856 [==============================] - 0s 181us/sample - loss: 0.5230 - acc: 0.8318 - val_loss: 0.4890 - val_acc: 0.8326
Epoch 53/150
856/856 [==============================] - 0s 201us/sample - loss: 0.5245 - acc: 0.8294 - val_loss: 0.4837 - val_acc: 0.8326
Epoch 54/150
856/856 [==============================] - 0s 187us/sample - loss: 0.5124 - acc: 0.8329 - val_loss: 0.4765 - val_acc: 0.8326
Epoch 55/150
856/856 [==============================] - 0s 181us/sample - loss: 0.5133 - acc: 0.8283 - val_loss: 0.4867 - val_acc: 0.8326
Epoch 56/150
856/856 [==============================] - 0s 179us/sample - loss: 0.5153 - acc: 0.8364 - val_loss: 0.4741 - val_acc: 0.8326
Epoch 57/150
856/856 [==============================] - 0s 169us/sample - loss: 0.5022 - acc: 0.8364 - val_loss: 0.4643 - val_acc: 0.8326
Epoch 58/150
856/856 [==============================] - 0s 259us/sample - loss: 0.5079 - acc: 0.8306 - val_loss: 0.4738 - val_acc: 0.8326
Epoch 59/150
856/856 [==============================] - 0s 290us/sample - loss: 0.5207 - acc: 0.8306 - val_loss: 0.4644 - val_acc: 0.8326
Epoch 60/150
856/856 [==============================] - 0s 243us/sample - loss: 0.5083 - acc: 0.8283 - val_loss: 0.4754 - val_acc: 0.8326
Epoch 61/150
856/856 [==============================] - 0s 181us/sample - loss: 0.5075 - acc: 0.8306 - val_loss: 0.4781 - val_acc: 0.8326
Epoch 62/150
856/856 [==============================] - 0s 172us/sample - loss: 0.5111 - acc: 0.8329 - val_loss: 0.4733 - val_acc: 0.8326
Epoch 63/150
856/856 [==============================] - 0s 179us/sample - loss: 0.5217 - acc: 0.8283 - val_loss: 0.4884 - val_acc: 0.8326
Epoch 64/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4984 - acc: 0.8353 - val_loss: 0.4877 - val_acc: 0.8326
Epoch 65/150
856/856 [==============================] - 0s 171us/sample - loss: 0.5074 - acc: 0.8341 - val_loss: 0.4807 - val_acc: 0.8326
Epoch 66/150
856/856 [==============================] - 0s 178us/sample - loss: 0.5022 - acc: 0.8318 - val_loss: 0.4764 - val_acc: 0.8326
Epoch 67/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4846 - acc: 0.8364 - val_loss: 0.4598 - val_acc: 0.8326
Epoch 68/150
856/856 [==============================] - 0s 232us/sample - loss: 0.5005 - acc: 0.8283 - val_loss: 0.4678 - val_acc: 0.8326
Epoch 69/150
856/856 [==============================] - 0s 182us/sample - loss: 0.5062 - acc: 0.8329 - val_loss: 0.4604 - val_acc: 0.8326
Epoch 70/150
856/856 [==============================] - 0s 173us/sample - loss: 0.5038 - acc: 0.8283 - val_loss: 0.4753 - val_acc: 0.8326
Epoch 71/150
856/856 [==============================] - 0s 186us/sample - loss: 0.5025 - acc: 0.8236 - val_loss: 0.4676 - val_acc: 0.8326
Epoch 72/150
856/856 [==============================] - 0s 254us/sample - loss: 0.4910 - acc: 0.8329 - val_loss: 0.4651 - val_acc: 0.8326
Epoch 73/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4988 - acc: 0.8306 - val_loss: 0.4656 - val_acc: 0.8326
Epoch 74/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4940 - acc: 0.8294 - val_loss: 0.4556 - val_acc: 0.8326
Epoch 75/150
856/856 [==============================] - 0s 213us/sample - loss: 0.5080 - acc: 0.8259 - val_loss: 0.4729 - val_acc: 0.8326
Epoch 76/150
856/856 [==============================] - 0s 175us/sample - loss: 0.4986 - acc: 0.8329 - val_loss: 0.4505 - val_acc: 0.8326
Epoch 77/150
856/856 [==============================] - 0s 174us/sample - loss: 0.5105 - acc: 0.8306 - val_loss: 0.4565 - val_acc: 0.8326
Epoch 78/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4963 - acc: 0.8318 - val_loss: 0.4640 - val_acc: 0.8326
Epoch 79/150
856/856 [==============================] - 0s 208us/sample - loss: 0.4874 - acc: 0.8364 - val_loss: 0.4654 - val_acc: 0.8326
Epoch 80/150
856/856 [==============================] - 0s 192us/sample - loss: 0.5014 - acc: 0.8294 - val_loss: 0.4708 - val_acc: 0.8326
Epoch 81/150
856/856 [==============================] - 0s 178us/sample - loss: 0.4992 - acc: 0.8341 - val_loss: 0.4709 - val_acc: 0.8326
Epoch 82/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4914 - acc: 0.8353 - val_loss: 0.4636 - val_acc: 0.8326
Epoch 83/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4843 - acc: 0.8435 - val_loss: 0.4800 - val_acc: 0.8326
Epoch 84/150
856/856 [==============================] - 0s 176us/sample - loss: 0.4985 - acc: 0.8318 - val_loss: 0.4568 - val_acc: 0.8326
Epoch 85/150
856/856 [==============================] - 0s 181us/sample - loss: 0.5021 - acc: 0.8329 - val_loss: 0.4669 - val_acc: 0.8326
Epoch 86/150
856/856 [==============================] - 0s 174us/sample - loss: 0.5114 - acc: 0.8248 - val_loss: 0.4696 - val_acc: 0.8326
Epoch 87/150
856/856 [==============================] - 0s 189us/sample - loss: 0.4781 - acc: 0.8329 - val_loss: 0.4622 - val_acc: 0.8326
Epoch 88/150
856/856 [==============================] - 0s 218us/sample - loss: 0.4992 - acc: 0.8376 - val_loss: 0.4743 - val_acc: 0.8372
Epoch 89/150
856/856 [==============================] - 0s 216us/sample - loss: 0.4932 - acc: 0.8388 - val_loss: 0.4593 - val_acc: 0.8326
Epoch 90/150
856/856 [==============================] - 0s 199us/sample - loss: 0.4898 - acc: 0.8388 - val_loss: 0.4607 - val_acc: 0.8326
Epoch 91/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4882 - acc: 0.8400 - val_loss: 0.4533 - val_acc: 0.8326
Epoch 92/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4838 - acc: 0.8376 - val_loss: 0.4464 - val_acc: 0.8372
Epoch 93/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4894 - acc: 0.8376 - val_loss: 0.4666 - val_acc: 0.8326
Epoch 94/150
856/856 [==============================] - 0s 175us/sample - loss: 0.5010 - acc: 0.8329 - val_loss: 0.4452 - val_acc: 0.8326
Epoch 95/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4760 - acc: 0.8446 - val_loss: 0.4485 - val_acc: 0.8326
Epoch 96/150
856/856 [==============================] - 0s 199us/sample - loss: 0.4873 - acc: 0.8353 - val_loss: 0.4423 - val_acc: 0.8326
Epoch 97/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4932 - acc: 0.8329 - val_loss: 0.4579 - val_acc: 0.8326
Epoch 98/150
856/856 [==============================] - 0s 171us/sample - loss: 0.4798 - acc: 0.8294 - val_loss: 0.4655 - val_acc: 0.8326
Epoch 99/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4864 - acc: 0.8364 - val_loss: 0.4450 - val_acc: 0.8326
Epoch 100/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4853 - acc: 0.8376 - val_loss: 0.4459 - val_acc: 0.8326
Epoch 101/150
856/856 [==============================] - 0s 182us/sample - loss: 0.4834 - acc: 0.8306 - val_loss: 0.4462 - val_acc: 0.8372
Epoch 102/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4985 - acc: 0.8353 - val_loss: 0.4570 - val_acc: 0.8326
Epoch 103/150
856/856 [==============================] - 0s 290us/sample - loss: 0.4786 - acc: 0.8400 - val_loss: 0.4452 - val_acc: 0.8326
Epoch 104/150
856/856 [==============================] - 0s 176us/sample - loss: 0.4934 - acc: 0.8353 - val_loss: 0.4517 - val_acc: 0.8326
Epoch 105/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4874 - acc: 0.8364 - val_loss: 0.4585 - val_acc: 0.8326
Epoch 106/150
856/856 [==============================] - 0s 189us/sample - loss: 0.4982 - acc: 0.8329 - val_loss: 0.4799 - val_acc: 0.8326
Epoch 107/150
856/856 [==============================] - 0s 201us/sample - loss: 0.4842 - acc: 0.8341 - val_loss: 0.4643 - val_acc: 0.8326
Epoch 108/150
856/856 [==============================] - 0s 203us/sample - loss: 0.4939 - acc: 0.8318 - val_loss: 0.4587 - val_acc: 0.8326
Epoch 109/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4853 - acc: 0.8400 - val_loss: 0.4550 - val_acc: 0.8326
Epoch 110/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4984 - acc: 0.8353 - val_loss: 0.4653 - val_acc: 0.8326
Epoch 111/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4979 - acc: 0.8306 - val_loss: 0.4768 - val_acc: 0.8279
Epoch 112/150
856/856 [==============================] - 0s 187us/sample - loss: 0.4962 - acc: 0.8341 - val_loss: 0.4509 - val_acc: 0.8326
Epoch 113/150
856/856 [==============================] - 0s 192us/sample - loss: 0.4750 - acc: 0.8446 - val_loss: 0.4486 - val_acc: 0.8326
Epoch 114/150
856/856 [==============================] - 0s 199us/sample - loss: 0.4888 - acc: 0.8353 - val_loss: 0.4517 - val_acc: 0.8326
Epoch 115/150
856/856 [==============================] - 0s 197us/sample - loss: 0.4848 - acc: 0.8400 - val_loss: 0.4671 - val_acc: 0.8326
Epoch 116/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4761 - acc: 0.8376 - val_loss: 0.4686 - val_acc: 0.8419
Epoch 117/150
856/856 [==============================] - 0s 182us/sample - loss: 0.5148 - acc: 0.8259 - val_loss: 0.4476 - val_acc: 0.8326
Epoch 118/150
856/856 [==============================] - 0s 189us/sample - loss: 0.4896 - acc: 0.8294 - val_loss: 0.4829 - val_acc: 0.8326
Epoch 119/150
856/856 [==============================] - 0s 168us/sample - loss: 0.4958 - acc: 0.8388 - val_loss: 0.4825 - val_acc: 0.8326
Epoch 120/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4922 - acc: 0.8400 - val_loss: 0.4656 - val_acc: 0.8372
Epoch 121/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4956 - acc: 0.8259 - val_loss: 0.4766 - val_acc: 0.8326
Epoch 122/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4774 - acc: 0.8446 - val_loss: 0.4546 - val_acc: 0.8326
Epoch 123/150
856/856 [==============================] - 0s 178us/sample - loss: 0.4838 - acc: 0.8423 - val_loss: 0.4498 - val_acc: 0.8326
Epoch 124/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4838 - acc: 0.8400 - val_loss: 0.4576 - val_acc: 0.8326
Epoch 125/150
856/856 [==============================] - 0s 187us/sample - loss: 0.4780 - acc: 0.8458 - val_loss: 0.4600 - val_acc: 0.8326
Epoch 126/150
856/856 [==============================] - 0s 181us/sample - loss: 0.4854 - acc: 0.8318 - val_loss: 0.4390 - val_acc: 0.8279
Epoch 127/150
856/856 [==============================] - 0s 178us/sample - loss: 0.4771 - acc: 0.8470 - val_loss: 0.4435 - val_acc: 0.8326
Epoch 128/150
856/856 [==============================] - 0s 213us/sample - loss: 0.4940 - acc: 0.8329 - val_loss: 0.4520 - val_acc: 0.8326
Epoch 129/150
856/856 [==============================] - 0s 190us/sample - loss: 0.4873 - acc: 0.8388 - val_loss: 0.4505 - val_acc: 0.8326
Epoch 130/150
856/856 [==============================] - 0s 178us/sample - loss: 0.4887 - acc: 0.8435 - val_loss: 0.4495 - val_acc: 0.8326
Epoch 131/150
856/856 [==============================] - 0s 175us/sample - loss: 0.4781 - acc: 0.8341 - val_loss: 0.4586 - val_acc: 0.8326
Epoch 132/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4935 - acc: 0.8329 - val_loss: 0.4487 - val_acc: 0.8326
Epoch 133/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4774 - acc: 0.8306 - val_loss: 0.4425 - val_acc: 0.8326
Epoch 134/150
856/856 [==============================] - 0s 187us/sample - loss: 0.4738 - acc: 0.8376 - val_loss: 0.4473 - val_acc: 0.8326
Epoch 135/150
856/856 [==============================] - 0s 242us/sample - loss: 0.4747 - acc: 0.8388 - val_loss: 0.4537 - val_acc: 0.8326
Epoch 136/150
856/856 [==============================] - 0s 180us/sample - loss: 0.5033 - acc: 0.8376 - val_loss: 0.4631 - val_acc: 0.8326
Epoch 137/150
856/856 [==============================] - 0s 183us/sample - loss: 0.4746 - acc: 0.8364 - val_loss: 0.4636 - val_acc: 0.8326
Epoch 138/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4681 - acc: 0.8481 - val_loss: 0.4366 - val_acc: 0.8326
Epoch 139/150
856/856 [==============================] - 0s 193us/sample - loss: 0.4691 - acc: 0.8435 - val_loss: 0.4447 - val_acc: 0.8372
Epoch 140/150
856/856 [==============================] - 0s 200us/sample - loss: 0.4810 - acc: 0.8376 - val_loss: 0.4403 - val_acc: 0.8326
Epoch 141/150
856/856 [==============================] - 0s 173us/sample - loss: 0.4791 - acc: 0.8306 - val_loss: 0.4359 - val_acc: 0.8326
Epoch 142/150
856/856 [==============================] - 0s 200us/sample - loss: 0.4785 - acc: 0.8388 - val_loss: 0.4355 - val_acc: 0.8326
Epoch 143/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4879 - acc: 0.8294 - val_loss: 0.4426 - val_acc: 0.8326
Epoch 144/150
856/856 [==============================] - 0s 180us/sample - loss: 0.4765 - acc: 0.8411 - val_loss: 0.4473 - val_acc: 0.8326
Epoch 145/150
856/856 [==============================] - 0s 192us/sample - loss: 0.4628 - acc: 0.8446 - val_loss: 0.4467 - val_acc: 0.8326
Epoch 146/150
856/856 [==============================] - 0s 186us/sample - loss: 0.4795 - acc: 0.8318 - val_loss: 0.4558 - val_acc: 0.8326
Epoch 147/150
856/856 [==============================] - 0s 172us/sample - loss: 0.4772 - acc: 0.8353 - val_loss: 0.4514 - val_acc: 0.8372
Epoch 148/150
856/856 [==============================] - 0s 178us/sample - loss: 0.4843 - acc: 0.8364 - val_loss: 0.4453 - val_acc: 0.8326
Epoch 149/150
856/856 [==============================] - 0s 179us/sample - loss: 0.4652 - acc: 0.8376 - val_loss: 0.4437 - val_acc: 0.8372
Epoch 150/150
856/856 [==============================] - 0s 195us/sample - loss: 0.4708 - acc: 0.8388 - val_loss: 0.4371 - val_acc: 0.8326
    In [416]:
    y_pred = model1.predict_classes(X_test)
y_classes = [np.argmax(y, axis=None, out=None) for y in y_test]
print(classification_report(y_pred,y_classes))

                  precision    recall  f1-score   support

           1       0.00      0.00      0.00         0
           2       1.00      0.81      0.90       528
           3       0.00      0.00      0.00         0

    accuracy                           0.81       528
   macro avg       0.33      0.27      0.30       528
weighted avg       1.00      0.81      0.90       528

    Final Conclusion:

  • We obtained better results overall after we boradly categorised the signal strength into 3 classes.
  • Model trained on PC has better overall weighted average f1score and it also tries to predict class 3 or stronger signals.
  • From above recall scores of all the classes , it is evident that we need some more data in order to increase the predictability of our model.

    • In [417]:
    model.save("CatModelTrainedOnPC.h5")
    In [ ]: